Table of contents

The 2020 5th International Conference on Intelligent Computing and Signal Processing (ICSP 2020) was successfully held on March 20-22, 2020. In order to comply with the control and prevention of the epidemic COVID-19, ICSP 2020 was adjusted to be held via online platform to avoid gathering and physical contact. Despite the influence of the epidemic, ICSP 2020 provides a flexible platform for the experts, scholars and practitioners in the field of intelligent computing and signal processing to exchange new ideas and present the state-of-the-art research. This conference also promotes the establishing of business or research relations among global partners for future collaboration. We hope that this conference could make significant contribution to the update of knowledge about these latest scientific field.

In this conference, we were honored to have over 100 experts and scholars in the area of Intelligent Computing and Signal Processing representing different famous universities and institutes around the globe to form our organizing committee and technology program committee. During the conference, we were pleased to invite three distinguished experts to present their insightful speeches. Assoc. Prof. Ata Jahangir Moshayedi, from Jiangxi University of Science and Technology, China, held a speech on Plume Tracking ROBOTS Missions and Obstacles in Robotic World. Assoc. Prof. Dr. Nada M. Al Hakkak, representing Baghdad College of Economic Sciences University, IRAQ, talked about Resource Allocation in Computer Simulation. And Prof. Hsiang-Chen Hsu, from Department of Industrial Management, I-Shou university, shared a speech on his specialty related to Reliability Analysis for IC Package.

List of Committee member, Organizing Committee, Technology Program Committee are available in this pdf.

All papers published in this volume of Journal of Physics: Conference Series have been peer reviewed through processes administered by the proceedings Editors. Reviews were conducted by expert referees to the professional and scientific standards expected of a proceedings journal published by IOP Publishing.

This is an introductory paper based on general knowledge such as simplicial complex, nerves and their homotopy equivalents in the topology field. It focuses on the theoretical part of one of the four statements on nerve theorem of paper variations on the nerve theorem. One of many ideologies throughout this theorem is to conduct an approximating process on mathematical spaces with restrictions, which is the bridge from a theory to an application of Weil's theorem and many others. Besides, work done by Kathryn Heal in her paper variations on the nerve theorem has proved the correctness and feasibility about Weil's theorem. This paper aims to clarify her proof of Weil's theorem and make it more easier for people to understand.

In the process of learning the data structure, it is very necessary to master the sorting algorithm, and in the program design, the sorting algorithm is applied frequently. Based on the importance of sorting algorithms, this paper will carefully compare the characteristics of different algorithms, starting with the work efficiency, algorithm implementation, basic ideas, sorting methods and other aspects, and draw conclusions so as to better select sorting algorithms.

This article analyzes the basic classification of machine learning, including supervised learning, unsupervised learning, and reinforcement learning. It combines analysis on common algorithms in machine learning, such as decision tree algorithm, random forest algorithm, artificial neural network algorithm, SVM algorithm, Boosting and Bagging algorithm, BP algorithm. Through the development of theoretical systems, further improvement of autonomous learning capabilities, the integration of multiple digital technologies, and the promotion of personalized custom services, the purpose is to improve people's awareness of machine learning and accelerate the speed of popularization of machine learning.

Among the exploration methods of oil and gas resources, one of the most important exploration methods is seismic exploration. Due to economic development, the previous shallow exploration has been unable to meet the needs, and it is necessary to conduct exploration in deep geological layers. However, as the depth increases, the signal of the geophone gradually weakens and the noise interference increases, which makes the denoising work of seismic exploration more important. Aiming at the above problems, this paper proposes the research of seismic exploration equipment based on time-frequency space algorithm, and uses different window lengths for processing based on the time-frequency peak filtering algorithm. Through simulation experiments on seismic exploration data with different frequencies, the results show that the method proposed in this paper improves the effective signal amplitude maintenance by more than 11%, and the amplitude maintenance is higher than 88%. The signal-to-noise ratio also improved from -8dB before filtering to 12.13dB. This shows that the method proposed in this paper can better meet the needs of seismic exploration signal processing.

Inertial navigation system (INS) has high short-term accuracy, while satellite navigation positioning system (GNSS) has high long-term accuracy. The integrated navigation system integrates two or more different navigation systems with hardware and software, and uses their respective advantages to make up for the shortcomings of a single system. In this paper, a dual-antenna GNSS/MIMU integrated navigation system is designed, and experimental research is performed on different integrated navigation systems. Algorithms of GNSS satellite positioning and dual-antenna attitude measurement, strapdown inertial navigation attitude calculation and update, and integrated navigation filtering are studied. Static and dynamic experiments are performed to study the performance of the system under different GNSS/MIMU combinations. The experimental results show that the combined system has smaller errors in terms of horizontal position, heading angle, pitch angle, and roll angle.

The massive uncertain data generated by network applications has potential value, and it is of great significance to carry out clustering analysis of uncertain data. However, the uncertainty of data brings a serious challenge to traditional clustering algorithms. There are some problems in the existing clustering algorithms for uncertain data. (1) Some algorithms have a lot of meaningless distance calculations when calculating the distance of uncertain objects, and then the calculation complexity of algorithms increases. (2) The data model in clustering algorithms results in the loss of data distribution information, and then the accuracy of the algorithms decreases. In this paper, we propose an uncertain tuple density clustering (UTDC) algorithm for uncertain data. Firstly, we extract the data distribution features of uncertain instances and construct uncertain tuples by introducing the cloud model, which realizes the pruning of clustering objects. Secondly, we apply the EW distance to the traditional density clustering algorithm DBSCAN, which completes the density clustering for uncertain data. The experimental results show that comparing with UK-Means and FDBSCAN, UTDC algorithm effectively reduces the computational complexity and improves the accuracy.

A university campus has a high population density. When the construction workers repairs and reconstructs underground pipelines, they must complete the construction as soon as possible to reduce the impact on the work and study of teachers and students. The underground pipeline 3D information system can display the spatial location and attributes of the underground pipeline intuitively, providing the construction workers with operational guidance to ensure that the repair is carried out in an orderly manner.

Taking the cross-border e-commerce policy texts introduced in the past five years as the research object, the co-word clustering method is used to divide the policy texts into four categories, and the content is analyzed based on the theory of the ecosystem. The study found that Gansu Province currently has unbalanced cross-border e-commerce policy structures, insufficient specific policies issued by prefecture-level city governments, and weak linkages among government departments in issuing policies. Therefore, the relevant departments need to further improve the cross-border e-commerce policy in response to the above issues, so as to promote the high-quality development of cross-border e-commerce in Gansu Province.

This paper proposed a real time autocorrelation and limited sampling algorithm with fixed packet length based on Field Programmable Gate Array(FPGA) for the period estimation and significant data transmission for arbitrary periodic signal, which contributes to information transfer in the Internet of things for smart sensors, and has a wide applications such as radar, communication, sensors network and measurement. Firstly, an explanation of the algorithm is shown. Secondly, several simulation results are analyzed. Finally, experimental and hardware implementation results are shown.

At present, prevalent algorithms on parameters estimation of linear frequency modulation (LFM) signal include the fractional Fourier transform (FrFT) method, of which the crucial procedure is to estimate the optimal rotation angle. The estimated value of the optimal rotation angle has a great influence on the estimation accuracy of LFM parameters, and it also affects the computational complexity of the algorithm. This paper proposes an algorithm on parameters estimation of LFM signal with high efficiency and precision. Based on the mathematical relationship between the optimal and any two rotation angles of FrFT, the optimal rotation angle is calculated through three FrFT transforms and interpolation, and the LFM signal parameters are estimated with high efficiency and precision. Simulation results verify the effectiveness of the proposed algorithm.

In this paper, a set of dynamic handling control system based on Fanuc robot is designed and studied, in which the conveyor line positioning mechanism, industrial robot and stability mechanism are connected together by Siemens 1500 PLC as the main control unit. The machine vision and Fanuc robot tracking function are used to make dynamic visual photographs of the robot, which compensates for the deviation of the hanging rod and increases the success rate of the moving product. The monotonous and repeated manual labor can be solved, and the production efficiency and production quality of the enterprise can be improved by replacing the artificial handling with the robot handling.

With constraints in single image, such as collinear equations and coplanarity, a real-scale object model can be obtained, and real scenes of regular objects can be reconstructed, by using three-dimensional reconstruction algorithms such as grid processing and interpolation fitting. A 3D reconstruction algorithm based on single pipe weld seam ray picture is proposed. Firstly, recognizing and segmenting the weld image, and using non-linear filtering to enhance the gray level of the image. Secondly, identifying the inner and outer edge positions of the weld, and performing threshold segmentation, morphological filtering, connected area detection, edge detection, line detection, image rotation, and inner wall area extraction. Thirdly, based on the image of the elliptical opening of the weld seam, the inverse perspective transformation is used to restore the formation of the weld seam root and the weld 3D scatter points. Finally, 3D point cloud is meshed, the function of point-to-surface is achieved, and a three-dimensional model of the welding tumor is generated. The experimental results demonstrate that the algorithm can accurately and effectively perform 3D reconstruction of the weld seam based on a single 2D pipe radiograph of the weld seam, and realize accurate identification of the weld nodules at the root of the pipe through-flow weld.

The Beidou B1C signal is a new type of navigation signal used by the Beidou III system. Compared with traditional satellite navigation signals, this navigation signal has two significant characteristics: one is the use of BOC(Binary-Offset-Carrier) modulation, and the other is the separation of the data channel and the pilot channel. BOC modulation has multi-peak characteristics, which leads to ambiguity in tracking, while the PCF(Pseudo-Correlation-Function) algorithm can eliminate related secondary peaks of the BOC signal; the separation of the data channel and the pilot channel results in the dispersion of signal energy. Therefore, based on the PCF algorithm, this paper proposes a linear joint tracking algorithm suitable for B1C signals from the perspective of engineering applications. Finally, it is verified that the tracking accuracy of the proposed algorithm is better than the single-component PCF tracking algorithm.

Data imbalance increases the difficulty of knowledge mining from data. Aiming at the problem of data imbalanced classification, a random forest algorithm based on GAN is proposed, which can effectively classify imbalanced data sets. The GAN model is used to generate a few class samples, which are mixed with the original data samples to form a new data set to reduce the imbalance of data. Then the new data set is divided into several data subsets with a balanced sample distribution. Finally, all the decision trees are collected to form the forest and the classification results are obtained. The random forest algorithm based on GAN improves the classification effect on minority classes, which makes the equilibrium rate of the dataset reach 50%. The algorithm in this paper has performed relevant experiments on multiple imbalanced datasets, the results show that the algorithm has good performance in data imbalance classification. The optimized algorithm after parallelization improves the running speed.

In this paper, we focus on the diffraction problem of periodic strip gratings. A least-squares non-polynomial finite element method is proposed for this problem. Firstly, the computational domain is decomposed. Secondly, a local approximation space is defined by using non-polynomial functions in every subdomain. Then the least-squares finite element method is derived. Finally, numerical results are reported to show the effectiveness and convergence of the least-squares non-polynomial finite element method.

As a classical and significant problem, handwritten character recognition has been widely used in our daily lives. With recent deep learning methods, previous studies have achieved a great improvement for this problem in the past few years. However, the handwritten character recognition for Urdu, which is one of the largest languages of the world, is less studied in the existing literature. In this paper, we fill in this gap and evaluate different deep learning models on the problem of Urdu handwritten characters recognition based on a newly released dataset. Combined with data augmentation and transfer learning techniques, we achieve the state-of-the-art results by recognizing digits and characters with an accuracy of 98.94% and 99.08%, respectively, which greatly improves the baselines of 97% and 86.5%.

Aiming at the orthodontic path planning problems, for example, the number is estimated by orthodontists based on experience, and the efficiency of tooth movement path planning on computer is low. This paper proposed a teeth movement path planning method based optimized artificial bee colony algorithm (ABC). At first, according to the age and gender of patients, the constraint amount of tooth movement is automatically determined. Second, by optimizing the observation bees select method, using both roulette and anti-roulette mechanisms replaced roulette mechanisms, this method can avoid the population from falling into local optimality and obtain better distribution. Then, the initial stage number is obtained through interpolation, and combined with the initial food source generation strategy in ABC. Finally, this paper uses the optimized ABC to plan the teeth movement path and obtain the actual amount of orthodontic stages. The experimental result shows that the optimized ABC has better distribution than basic ABC, and the proposed method is closer to clinical orthodontic, reduced the workload of the orthodontist. The data shows this method is more suitable for different patient situations, reduced the orthodontic stage number and the cost of teeth movement effectively.

The essence of attribute reduction is to delete unnecessary or unimportant attributes in the knowledge base while keeping the classification ability of the information system. Most of the existing reduction algorithms are designed for static decision tables. However, when the data set changes dynamically, the time consumption of static reduction algorithms is huge and the efficiency is not high. At present, there are few studies on the reduction algorithms of dynamic decision tables. In this paper, an existing binary distinguishable matrix reduction algorithm is optimized by using the method of binary distinguishable matrix when the objects are dynamically increased. In the case of static data set, the algorithm is improved to deal with dynamic data set by judging the relationship between new data and original data set. Examples show that the algorithm is simple, efficient, accurate, and has certain practicability and completeness.

In order to solve the problem of odometer small differential error and pulse calculation error and improve the precision of SINS and Odometer Integrated Navigation, this paper proposes an error correction algorithm of vehicle SINS and odometer integrated navigation, compares the calculated position of SINS with the calculated position of odometer as the measured value of integrated navigation within the fixed mileage range and presents a method to correct the truncation error of odometer counting. The simulation results show that these methods can quickly separate the sensor errors of the integrated navigation system, have the high positioning and orientation accuracy. Also, these methods can be applied in the practical vehicle SINS system.

In recent years, with the development of cloud computing, virtualization technology has received widespread attention. As a new representative of virtualization technology, containers have been widely used in software development, operation and maintenance, testing and other aspects, such as microservices and Docker Cloud. In cloud data centers, containers have gradually replaced virtual machines (VMs) as a new carrier for cloud tasks. However, with the increasing number of cloud products, the scale of tasks requested by users in cloud data centers continues to expand. The economic cost of tasks in the process of containerized deployment has become a concern of various cloud service vendors. The container deployment cost usually includes the data exchange cost, the image pull cost and the server energy cost. In order to solve the containerized deployment of application tasks in the container cloud environment with the lowest possible container deployment cost, this paper proposes a new cost calculation model in a container cloud environment, and then presents an improved particles swarm algorithm, namely a particle swarm optimization (PSO) algorithm for container deployment (CD-PSO) to provide the best solution for application task loading. Experimental results show that the proposed algorithm has a lower deployment cost than other scheduling algorithms.

With its strong search ability and other characteristics, the genetic algorithm in mobile robot path optimization has been rapidly developed and applied, and has become one of the most successful algorithms to solve the mobile robot path planning problem. Firstly, this paper introduces the basic principle of genetic algorithm and gives a brief overview of its development and improvement strategies. Then, combining several representative improvement strategies, path planning of mobile robot is explained in detail. Finally, combining the theoretical and applied research results of genetic algorithm, the further development of the algorithm is summarized and prospected.

With the rapid development of unmanned surface vessels (USV) applications, USV are widely used in military intelligence collection, target monitoring, and shipping services. However, this also brings security issues to shipping and the country. For more efficient detection of USV, in this paper, an enclosing center distance Intersection over Union (E-CIoU) algorithm is proposed for real-time USV detection, where the normalized distance between center points of the smallest enclosing box and target box is combined with bounding box regression. Meanwhile, we designed a new USV data set. The extensive experiments on USV data set have demonstrated that the proposed approach has achieved better convergence speed during training and a significant accuracy prediction.

This paper proposes a combined cell voltage acquisition algorithm based on Kalman filtering, which is used to solve the problem of insufficient cell voltage acquisition and high delay in the current aluminum electrolysis process. This algorithm is based on Kalman filtering, and uses the mean square error of the data to represent the Gaussian white noise power of the collected and predicted values. It has strong tracking performance in the state of stable voltage. The voltage acquisition data out of the interval is comply to first-order inertial filtering, to replace the voltage sampling value, and then using the Kalman filtering algorithm to calculate the filtering gain, to ensure that the combined algorithm can filter out the influence of noise when the cell voltage fluctuates greatly, and in the stable process of electrolysis, the convergence of battery voltage can be tracked quickly. The simulation results show that compared with the first-order inertial filter, the improved Kalman filter reduces the root mean square error by 50% when the voltage is stable, and reduces the convergence time by 50% after the electrolytic cell regenerates needle vibration and swing.

The nominal group technique (NGT) is a pattern of group decision-making technique in organizational behavior. Its research work is mainly focused on applications in different fields, and there is less research on algorithm simulation. In this paper, a method of NGTPSO is proposed in which NGT will be introduced into PSO and an attempt will be made to simulate NGT using PSO. From the perspective of PSO, PSO using NGT adds group decision-making technique. It characterizes the activity of individual and group through the grouping, individual optimizing, small-group discussion operation, large-group discussion operation and large-group decision operation. From the perspective of NGT, the decision process of NGT is simulated by PSO. The results show that the simulations of the algorithm are effective. The simulations show that small-group decision is better than or equal to individual decision and large-group decision is better than or equal to small-group decision. This is consistent with the basic principles of organizational behavior and it provides a good explanation for the relationship between NGT and PSO.

For dense individual targets, their close distances may cause gates overlap, which impairs the tracking performance. To solve this problem, this paper proposes a group target tracking algorithm based on topology. Firstly, the targets are grouped based on their positions and velocities. Note that group targets have more stable tracks compared with individual targets. Then, for the targets in the group, data association based on topology for dense individual targets is achieved, considering relative positions of targets are stable. Finally, the simulation experiments are conducted to evaluate the proposed algorithm. The simulation results show that the group target tracking algorithm based on topology improves the association performance compared with the traditional Hungarian algorithm.

Automotive instrument detection is usually performed manually. Factors such as human subjectivity and external environment have a great influence and are prone to errors. This paper proposes a method for automatic detection of pointer automobile instruments based on machine vision. Firstly, the common part of multiple instrument images with different pointer angles is extracted as the background image; secondly, based on the frame difference method, binarization, morphological operation and other techniques the instrument pointer is identified, and then the intersections of the pointer extension lines are used to determine the center point; thirdly, we extract the image of the scale position and calculate the scale template array; finally, the pointer value of each input instrument image is calculated based on the scale template array. Experiments show that this method can effectively solve the shortcomings of poor applicability and poor intelligence of existing algorithms, and can identify the value of the pointer accurately.

Abnormal events refer to specific events, such as forest fire, occurring in the wireless sensor networks of the Internet of Things (IoT), whose behaviors are quite different from normal events. By learning the underlying structure of sensor data, users can be helped to learn about the occurrence of these events as soon as possible. Due to the large number of sensors in the IoT and the periodic collections of data, sensor data has the problems of high dimensions and large amount, and the transmission of a large amount of data in the network is not a small challenge for bandwidth. In addition, the sensor data is unlabeled, so it is time-consuming and unrealistic to manually label all the data. Abnormal events in the IoT require low delay, such as gas concentration monitoring, In the IoT, data is generated continuously, so a well-trained model cannot remain unchanged, and features of new data need to be continuously learned. On account of the limitation of hardware, edge nodes cannot undertake the complicated and time-consuming task of model training and detection. None of the existing algorithms can meet the above requirements well, so this paper proposed an algorithm based on Compressed Sensing and Online Extreme Learning Machine Autoencoder named COELMAE. The proposed algorithm can carry out anomaly detection in low delay, unsupervised and online learning, and also has low computational complexity. What's more, the algorithm can reduce the amount of data transferred about 60%.

The quality of water is a key factor that determines whether aquaculture will be successful or not. Currently, aquaculture is mainly based on worker's experience, which is not capable of precisely controlling the quality of the aquaculture water. The main function of the existing system for aquaculture water quality is still to manage water quality, but hardly to analyse or visualize the data of water quality. Furthermore, fewer works for prediction of future water quality are made. To tackle these issues, we design an analysis and decision-making system based on deep learning for aquaculture water quality. We combine the latest artificial intelligence and big data technology to achieve the functions of the system such as the visualization of data, the prediction of future water quality factors and early-warning of water quality, which helps the users to realize scientific aquaculture.

With the growth of user popularity of online music platforms, the amount of music data and the demands on accurate recommendation system for music also increases. Therefore, an accurate and efficient recommendation system is urgently needed. In this paper, a music recommendation system based on fusion deep learning models is proposed. Both of the audio information and lyrics data are originally exploited. A supervised training method is applied to train this model with annotated data. In this paper, the result of the test dataset shows that the proposed model can achieve more than 90% accuracy performance on average.

Capsule networks (CapsNets) is a commonly used neural network structure, especially in computer vision research., which may have a significant impact on deep learning. Convolutional neural networks (CNNs) have reached superhuman level in various computer classification tasks such as classification, parsing, object detection, semantic segmentation, and instance segmentation. Traditional deep learning tends to deteriorate when the relevance of detection objects changes. For this kind of situation, this paper improves a new network, namely the capsule network(CapsNets), the network can achieve a higher detection rate in the hostile environment.

with the Rapid Development of Science and Technology in China, Modern Information Technology Has Been Deepened and Popularized in China. At the Same Time, It Has Also Developed to a Certain Extent in the Classroom Teaching of Primary School Language. At the Same Time, the Education Reform Has Made Great Changes in the Teaching Methods of Education, and the Teaching Requirements for Each Stage Have Gradually Improved, Especially the Pupils Who Are Extremely Malleable. At the Same Time, Because of the Development of Science and Technology, Teachers Gradually Found Ways to Improve the Efficiency of Primary School Teaching: Traditional Education Combined with Modern Technology. the Development of Technology Has an Ever-Increasing Impact on Society. Therefore, in Primary School Teaching, Multimedia Technology and Traditional Media Teaching Are Combined.

the Task of Robot Path Planning Based on Artificial Intelligence Algorithm is to Determine an Optimal Strategy, with the Goal of Maximizing the Return. When the Robot is in a New State, It Needs to Select an Action from the Action Set and Execute It. in Order to Improve the Quality of the Algorithm, It is Necessary to Apply the Search Strategy to the Action Selection. in the Path Planning System of the Robot, One of the Most Critical Links is the Realization of Directional Reference Vehicle Scheduling. Generally, Artificial Intelligence Algorithm is the Most Widely Used in Robot Path Planning. in This Paper, an Artificial Intelligence Algorithm is Used to Optimize the Vehicle Scheduling Problem. the Path Variable of the Problem Can Be Defined Correspondingly, Which Originates from the Planning Node. the Sequence of Sending Points Reached by Completing Other Sending Requirements is Called the Legitimate Sub Path, While the Corresponding Traversal Path Does Not Include the Sequence of Several Legitimate Sub Paths of Repeated Planning Points. the Algorithm Can Significantly Improve the Efficiency of Multi-Robot System, Reduce the Number of Explorations and Speed Up the Process of Convergence.

Nowadays, object detection has gradually become a quite popular field. From the traditional methods to the methods used at this stage, object detection technology has made great progress, and is still continuously developing and innovating. This paper reviews two-stage object detection algorithms used at this stage, explaining in detail the working principles of Faster R-CNN, R-FCN, FPN, and Casecade R-CNN and analyzing the similarities and differences between these four two-stage object detection algorithms. Then we used HSRC2016 ship dataset to perform experiments with Faster R-CNN, R-FCN, FPN, and Casecade R-CNN and compared the effectiveness of them with experimental results.

A multi-layered non-volatile solid-state reflective display pixel structure composed of a Fabry-Perot type resonance cavity and an ultrathin phase change material film has been fabricated and analysed. By electrically switching the phase states of a phase-changing Ge₂Sb₂Te₅ (GST) film using an electric micro-heater under the stack, a pronounced colour change in pixel structures is observed, which is attributed to the change in the resonance condition of the multi-layered stack. Furthermore, the range of attainable colour can be widened by exploiting a spatial colour mixing structure.

In the construction of digital library, it is difficult to find the content information of massive electronic books.Thus, the problem of many islands of information is presented, which hinders the efficient utilization of resources. This paper proposed an ontology framework of the electronic book domain to realize the semantic description of information resources. The KeyGraph algorithm was used to extract the key words of the book content abstract and chapter content of the electronic books, and the corresponding classes in the domain ontology were instantiated with the keyword content. Finally the semantic retrieval model has been proposed for extracting the query keywords. Content abstract and chapter content were obtained from book ontology. Compare query keywords with their similarity to get the final query result.

Electronic voting is the product of the development of modern network technology, in view of the electronic voting system security issues. This paper puts forward a kind of electronic voting scheme based on ElGamal homomorphism encryption, the scheme using homomorphic encryption strategies to voters vote operations, to protect the identity of voters and will not be exposed, to realize the security of electronic voting, confidentiality, and fairness, eligibility, and implement the Internet security electronic voting.

The DCM pile is applied in immersed tunnel foundation improvement first time at home in the mega project of Shenzhong Link with improved ground areas over 700000 m³. The compound foundation composed of gravel cushion and DCM piled foundation settlement law is not definite. To explore this law loading plate test in place on site is carried out and the test results show : (1)under the first order of load effect the settlement amount of the compound foundation accounts for over 60% of the total settlement, the modulus of compression of the gravel cushion is 1.5MPa being approximately 10%∼20% of normal modulus of compression; (2)under other loading effects the final deformation of compound foundation settlement curve shows stable state with the modulus of compression of the gravel cushion in range of 4.5∼12MPa which is equivalent to the normal value. The numerical analysis for purpose of the testing shows that the gravel cushion settlement accounts for larger proportion being 58.57% of the total settlement and the gravel cushion thickness, density, residual layers fallen in DCM dredging, all may be factors affecting settlement. Hence it is suggested to enhance study on the workmanship of the tunnel bed formation in order to improve the bed rigidity and anti-deformation ability.

The information management of engineering investigation quality is actually the process of producing, collecting, inputting, sharing and applying engineering investigation parameter data and management information. Using the theory of total quality management and the information technology such as blockchain, big data and cloud computing, this paper explores the possibility of constructing the information management system of engineering investigation quality from "data standard of engineering investigation quality information", "information management platform of engineering investigation project" and "geological information database of Engineering Investigation", and takes Chongqing as an example to carry out the engineering investigation quality. The construction and analysis of information management system verify the effectiveness of the information management system of engineering investigation quality.

An intelligent car plate detection method can make the travel more convenient and efficient. However, traditional methods are reasonably effective under the specific circumstances or strong assumptions only, and there are few databases for car plate detection. Therefore, a novel real-time car plate detection method based on improved Yolov3 has been proposed. In order to select the more precise number of candidate anchor boxed and aspect ratio dimensions, the K-Means algorithm is utilized. To solve the short of the available car plate database, a car plate database which has 6668 pictures has been established. As shown in the experimental results, the method which is proposed by this paper is better than original Yolov3. Thanks to the car plate database, the proposed method obtained better results even in the situation of inclination, too bright or too dark, different weather and so on.

In order to quantitatively characterize the uncertainty of hesitant fuzzy linguistic information, the entropy and similarity measures of hesitant fuzzy linguistic term sets are studied. First, the axiomatic definitions of entropy and similarity measures of hesitant fuzzy linguistic term sets are given, and general formulas of entropy and similarity measures for hesitant fuzzy linguistic term sets are proposed. Moreover, the algorithms for generating entropy and similarity of hesitant fuzzy linguistic term set are given. Then the relationship between entropy and similarity of hesitant fuzzy linguistic term set is studied, and the general formula of entropy based on similarity is proposed. These lay the theoretical foundation for flexible selection of entropy and similarity in multi-attribute decision making.

Under the implementation of the national strategy, "Internet +" is closely integrated with various traditional industries and has become one of the main driving forces for the economic development of the industry. For industrial development, "Internet +" provides technical tools, basic platforms, and link elements for building an ecosystem. The translation of commercial advertisements between English and Chinese is an important part of the language service industry. It is a major member of the cultural industry and guarantees the demand for commercial advertisement language services under the "Belt and Road" strategy in China. The combination of "Internet +" and commercial English-Chinese translation can promote the transformation and upgrading of commercial English-Chinese translation services. When translating commercial advertisements between English and Chinese, it is not only just a simple conversion of corresponding code symbols, but also needs to comprehensively consider the language background of national culture, customs and social development. In the context of "Internet +", the translation needs to reorganize the original information on the premise of ensuring that the semantics are basically equivalent and the functions are similar. With the help of modern computer network technology and cloud computing platform technology, this article builds an Internet-based translation platform to provide support for the translation, thereby achieving good translation results.

Software behavior research on behavioral statements is mostly software behavior analysis for specific software, and the characterization of software behavior is only obtained by simple buried point monitoring or static analysis. In this regard, this paper proposes a model for credibility analysis of most software behaviors based on the detection sandbox environment and the actual operating environment. The research developed a trusted plug-in based on the Kylin operating system platform. It adopted the idea of "consistent words and deeds", and used the actual behavior statement and the trusted behavior in the trusted behavior statement database for tested evaluation. In the aspect of software behavior tested measurement, a behavior statement judgment model based on cosine similarity algorithm is proposed to judge the reliability of software behavior. Experimental results: the trusted plug-in platform was able to perform credibility analysis on most software behaviors, and the introduced performance overhead was relatively low. It had a high ability to identify the credibility of software behaviors. It provides new ideas for studying the credibility of software behavior on the Kylin operating system platforms.

Approaches of machine translation are streamlined into three types based on strategies of knowledge processing, i.e. syntax-based methods, rule-based methods, and corpus-based methods, with their merits and demerits assessed. Consequently, the updated research tendency of the neural networks, as the current mainstream methods of machine translation, mainly comprising computational complexity reducing, words alignment enhancing, prior knowledge and constraints incorporating are stressed. Eventually, the future development orientations of the mainstream methods of neutral machine translation, particularly that of networks integration, data parallelizing and training are prospected.

The radar based human micro-motion research is a new technology rapidly developed in recent years. Radar detection of human target has advantages that other sensors do not have. The radar system is not affected by light or weather condition, which makes it possible to achieve detection in various environments. Besides, targets hidden behind obstacles can be detected, because electromagnetic waves can penetrate barriers of a certain thickness. Doppler frequency shift caused by the modulation of electromagnetic waves by the swing of hands and legs during human motion can be obtained by various signal processing methods. Separation and estimation of multi-components signals is a prerequisite for studying micro-motion characteristics of human target. Because there are a lot of aliasing in the time-frequency diagram of component with each other, separation of multi-components signals is a difficult but urgent problem at present. In addition, kinematic parameters of human legs play a key role in distinguishing human and animal targets. Therefore, in order to extract human leg signal and estimate its frequency accurately, a signal separation method based on Bezier curve fitting is proposed in this paper. Because human radar echo signals are often non-stationary time-varying signals, we need to process the collected signals with time-frequency transforms. The accuracy of this method for extracting components of legs is precise, and it has met the accuracy requirement of Micro-Doppler research. The simulation and experimental results prove the effectiveness of this method.

In view of the characteristics of computer simulation technology, the training program of professional talents is studied. Summarizing the problems existing in the training of professional skilled personnel, the aim is to perfect the method of cultivating professional skills in order to show the training value of computer simulation technology and provide reference for the development of the industry.

Conceptual semantic similarity calculation is an important content in research fields such as natural language processing, machine translation and is widely used in information retrieval, knowledge acquisition, text classification, and clustering. It's very important to development of natural language processing by improving the accuracy of semantic similarity calculation. This paper proposes a method for calculating the similarity of concepts based on the part-whole relationship between concepts. This method is based on the partial overall relationship diagram and uses the hierarchical features and positions of the concepts in the diagram to calculate the conceptual similarity.

The basis of multi-camera system measurement is the calibration of the multi-camera system. Although multi-camera system calibration has achieved certain results, there are still problems such as tedious calibration process and high calibration cost. In this paper, we propose a practical calibration method for multi-camera systems without overlapping fields of view. Firstly, a global 3D model of the scene is reconstructed using Simultaneous Localization and Mapping (SLAM). Then, each camera is calibrated using the 2D-3D correspondence between the images acquired by the multi-camera system and the 3D model. Finally, the external parameters of the multi-camera system are calibrated. In the process of using SLAM to build the model, the graph optimization is used to improve the accuracy of the modeling, and the accuracy of the calibration depends on the accuracy of the modeling. After several experiments, the experimental results show that the method is feasible.

With the rapid development of the Internet, online learning has developed rapidly, and learners are increasingly demanding the personalization and practicality of exercise. Facing the massive exercises in the online learning platform and the online examination system, how to choose the exercises that can be targeted and can make up for the knowledge loopholes has become a hot topic in the field of personalized recommendation of current teaching resources. In view of the fact that learners have a variety of learning features, and there are a large number of online exercises, various types and varying degrees of difficulty, this paper proposes a precise recommendation method based on Multidimensional features analysis for exercises. It quantifies the potential relationship between the learner and the exercise from three aspects: the heat of the exercise itself, the relevance of the knowledge among the exercises, and the similarity of the learner's style, using the linear combination and the Learning to Rank method to build the recommendation model to match learner and exercises exactly. Experiments show that the mean average precision of the ReMFA method reaches 36.8% when recommending five candidate exercises, which can provide learners with personalized exercise recommendation services, thereby improving the learner's learning efficiency.

At present, passive reconnaissance target location usually adopts multi-station direction-finding cross or multi-station time difference method, which has the disadvantages of complex scenes, large resource consumption, long time consumption and low positioning accuracy. In this paper, intelligence technology is used to perform machine learning and predictive evaluation of target radiated power in battlefields, and combined with the electromagnetic propagation model to complete the estimation of target approximate position of single-station passive reconnaissance. The results show that after training with a certain sample, the method can quickly complete the estimation of the target position, and has high positioning accuracy.

This paper proposes a network isolation method for large-scale cloud environments. The method run an Envoy-based endpoint program to proxy all network traffic on a physical node. And by configuring a route discovery registration center to uniformly maintain the network meta information of all virtual machines, the virtual machine uses the native VLAN technology to divide the user's subnet to ensure the communication performance of the subnet nodes and the virtual machine network maintained by the route discovery center. Meta information improves the speed of virtual machine ARP addressing and improves cross-node network request performance. Envoy's network proxy technology is used in conjunction with the route discovery center to broadcast the subnet broadcast communication of all virtual machines to the route discovery center, which improves the accuracy of the broadcast and greatly improves the utilization efficiency of the public network bandwidth.

In order to realize the effectiveness and scientificity of college physical training, we can integrate modern technology into college physical training, and promote the intellectualization and standardization of college physical training. In view of many problems that need to be solved in college physical education training, this paper designs a physical education training virtual simulation system by combining the application advantages of computer virtual reality technology, introduces the system functions and overall framework in detail, designs the corresponding functional modules, and expounds the advantages of computer virtual reality technology in college physical education training in detail. The research results show that it is necessary to apply computer virtual reality technology in college physical training, which can effectively promote the development of college physical training teaching, and promote college students to strengthen physical training.

Forest fire has always been an important hidden danger in forestry management. It is necessary to monitor and manage forest fire through video monitoring system. However, due to the lack of real-time performance of traditional video monitoring system and the difficulty in data processing, an embedded intelligent video analysis system is designed. The system uses a software and hardware design method based on Zynq SoC, first, each functional module of the hardware architecture is divided, then the system software is designed, and finally the system is evaluated as a whole. Based on the application requirements of forest fire prevention, this paper focuses on the detailed description of the underlying hardware architecture and operating system software design, and provides some value and reference for the application scenarios of forest fire prevention.

The deployment of many advanced pedestrian detection applications is largely hindered by the high computational cost of deep convolutional neural networks (CNNs). In this paper, we propose a two-step pruning method to design a lightweight pedestrian detection network in railway scenes. The first step is feature pyramid network (FPN) pruning, which utilizes the characteristic of pedestrian in railway scenes and the FPN structure in YOLOv3. The second step is regular channel pruning, which utilizes network slimming knowledge and is an accelerator-friendly pruning strategy. Our two-step pruning method gives about 88% reduction in parameters and about 74% reduction in computing complexity with comparable detection accuracy.

With the rapid development of Internet technology, colleges and universities have designed a music distance teaching system with the help of computer. It is not only convenient for students to study actively, but also uses modern information technology and Internet technology to carry out teaching activities. Moreover, the network-based autonomous learning can provide more free space for students. Therefore, the design of music distance teaching system has very important practical significance. In this paper, the principle of system design and the structure and function of the system are analyzed comprehensively, and the practicability and feasibility of the system are proved by the practice of the system, so it is worth popularizing and applying.

The aim of this study is to establish a reliable and widely applicable muscle strength (MS) estimation model based on the Mechanomyography (MMG). Seven healthy male volunteers were recruited to collect MMG and MS during the isometric contraction of their triceps. For MMG, 18 features were extracted. For the extreme gradient boosting (XGBoost) model and the quadratic polynomial (QP) model, the feature combination with the best estimation result was selected. The MS estimation performance of the XGBoost model and the QP model were compared. The performance of the QP model on the estimation of MS in different frequencies, different fatigue states and time periods was evaluated by using t-test. The results showed that when the number of features exceeds three, the model estimation accuracy has not improved significantly; and there was no significant difference in the estimation result of MS between the two models (p < 0.05), though the QP model was slightly better. The normalized root mean square error (NRMSE) and goodness of fit R of the MS estimation by the QP model were: 0.1343 ± 0.0296 and 0.8273 ± 0.0376. There was no significant difference in the MS estimation results in different conditions (p < 0.05).

Previous studies have proposed semantic-based approaches for code search over large-scale codebases, which has bridged the gap in understanding the semantics between natural language and source-code language. However, these studies either failed to determine an effective method for semantic representations or lacked the distinction of semantic features. In this study, we propose a novel attention-based LSTM neural network known as CSDA (Code Search based on Description Attention), which can effectively improve the code search performance. The proposed model can focus on different parts of a semantic feature when numerous aspects of a source code snippet are used as input. As opposed to assigning the same weight to different parts of the semantic vector, CSDA takes the semantics of natural language descriptions into account, so that the subtle differences hidden in the code snippet can be discriminated and associated with the corresponding queries. We compare CSDA with the existing state-of-the-art approach CODEnn, which uses a jointly embedding technique for code search. Our experimental evaluation demonstrates that CSDA outperforms previous methods and achieves superior code search performance to CODEnn, with higher success rates and mean reciprocal ranks. This study provides significant insights into the use of semantic representation methods in deep learning-based code search approaches.

A hardened computer is an army vehicle equipment. During the testing of electromagnetic compatibility, the RE102 electromagnetic radiation test item appeared multiple points exceeding the standard. In view of this problem, the electromagnetic environment of a reinforced computer is analyzed, and it is found that there are adverse factors that cause electromagnetic leakage in the aspects of the case aperture, cable shielding, power protection, and grounding methods. After optimization, the device passed the RE102 test smoothly. The radiated emission level of the improved equipment has decreased significantly, and the radiated disturbance suppression measures adopted are easy to implement and universal, which provides a reference for subsequent research and the development of similar equipment.

The increasing of car ownership makes the problem of parking difficult to become more and more prominent. The application of three-dimensional garage is one of the most effective solutions at present, which saves both cost and space. Based on the traditional three-dimensional garage, this paper designs a cylindrical three-dimensional garage system. The lifting device of the vehicle is composed of screw, parking platform, and transmission track. The control system of the garage is designed with the single-chip microcomputer as the control core. Finally, the control circuit of the software simulation system realizes the function of automatic exit storage of vehicles, which verifies the feasibility of the design.

Aiming at traditional equipment's problems, such as low tapping efficiency, poor reliability, a set of special automatic tapping machine control system is designed. The system can tap five workpieces at the same time by using Siemens S7-200 series CPU226 AC/DC/Relay PLC technology. The system will carry out technical reform of traditional equipment. The paper describes the control system scheme, work flow chat, I/O wiring diagram and program design of PLC. The operation results meet the use requirements. The system has the characteristics of good maneuverability, high operation safety and high production efficiency, which can greatly reduce the production cost.

Edge intelligent computing devices are often deployed in some extreme environments, where the transmission network bandwidth is low or the network environment changes greatly. Therefore, the traditional queue scheduling algorithms cannot guarantee the QoS of edge intelligent computing. WF²Q+ allocates bandwidth according to a fixed weight, which causes real-time data flow delay to increase when the network is unstable. The dynamic perception scheduling strategy proposed in this paper is to dynamically change the weight of WF²Q+ by dynamically sensing the backlog length of the queue. At the same time, combined with the queue scheduling algorithm of PQ, this algorithm can prioritize the transmission of real-time data with certain fairness. In addition, the token bucket algorithm is used to limit the sending rate of the device and prevent network congestion caused by burst data injection into the network. After experimental simulation, the improved algorithm can achieve good results on the delay index.

With the advancement of science and technology and the enhancement of human aesthetic experience, the fusion of technology and art has driven interactive installation art, an artistic expression with strong modern technology and information age characteristics, and has achieved unprecedented development. Processing is an open source programming language and development platform suitable for graphic and animation design. Through the study of processing, we have a more intuitive understanding of the creation of interactive art. Taking "what else" as an example, this paper studies the application value of Processing programming in interactive installation art, and discusses the influence of using code tools to innovate and develop art, integrating science and technology into art creation, aiming at providing new expression ideas for future installation art.

Against the background of a gradual increase in the elderly population, the care system for the elderly appears extremely important. The nursing system based on the cloud platform Internet of Things mainly includes the design of the lower computer, the debugging of the upper computer and the design of wireless communication. The design of the lower computer uses STM32, pulse detector, accelerometer, gyroscope and other modules. And with the cooperation of the host computer and the communication module, it can realize health monitoring functions such as medicine warehouse switch detection, elderly heart rate monitoring and fall detection. In addition, family members can also use the mobile phone APP to control connected devices around the elderly, such as smart pill boxes, to remind them to take medicine. There is no doubt that the design will greatly reduce family stress.

Analog electronic technology is a core basic course of electronic specialty in colleges and universities. It has the characteristics of complex theory, abstract content, difficult to explain by teachers and difficult to understand by students. This article mainly discusses some typical applications of Multisim virtual simulation software in the teaching of analog electronic technology courses, and analyzes the impact on classroom teaching methods. The practice shows that the application of simulation software in the teaching of simulation electronic technology can solve the problems of insufficient experimental teaching site and potential safety hazards. It can help students to understand abstract theory and master related knowledge. It can help to organize lively and efficient classroom teaching and promote the quality and efficiency of classroom teaching.

Needle insertion usually performed with the aid of an Ultrasound (US) image in the regional anaesthesia. Getting the position of the needle tip from US requires the technical expertise and skilled anaesthetist. To detect where the needle tip is around the nerve during the needle insertion, a modified block needle attached with a force sensor is used to feedback the amount of counter force from the needle tip. This paper is to investigate whether needle force under ultrasound guidance can guide anaesthesiologists in identifying the relative position of the needle tip and nerve The goal of our research was to use the visual needle tip force to guide the anaesthetist to detect the relative positions between the needle tip and nerve.

Traditional Chinese medicine (TCM) data is the main knowledge resource of TCM, which contains a wealth of clinical experience knowledge. Machine learning has made remarkable achievements in natural language processing. As the carrier of TCM knowledge and information stored in the form of text, using machine learning method to study these TCM data can save a lot of manpower cost, improve the objectivity of TCM, promote TCM related knowledge better, and have certain guiding significance for the research of TCM human engineering experiment. This paper proposes a recommendation algorithm based on mutual information clustering. Its core idea is calculating mutual information between two symptoms, and set symptom "relatives and friends group", after getting the symptom clustering results of mutual information, then combine the clustering results and search algorithm to achieve the effect of recommendation and filtering. Experimental results show that the proposed method is effective.

Milk powder is an important food, suitable for preservation and transportation. Protein is an important nutritional component in milk powder. At present, the routine physical and chemical analysis method is usually used in the detection of components of milk powder, which is time-consuming and labor-consuming. Therefore, it is very important to carry out the rapid non-destructive detection of milk powder quality. The application of near infrared spectroscopy (NIRS) technology in the rapid detection of milk powder quality is increasingly mature, but many analytical techniques are not perfect. In this study, the models of prediction protein in milk powder were established by R language with NIRS, and the characteristic bands were selected by recursive feature extraction (RFE) in R language, and the selected bands were screened one by one to determine their importance. Finally, using Partial least squares (PLS), generalize linear model (GLM), support vector machine (SVM), least angle regression (LARS), linear model (LM) and other methods to build the prediction models with 8 characteristic bands, the R-squared of the models can be increased by 0.1 to 0.7, the model robustness greatly improved.

In recent years, with the rapid development of the Internet in China, online transactions have grown greatly. For example, OTAs with a large number of hotels have accumulated a large amount of hotel data and user consumption data. And the online sales of hotels is the basis and core of revenue management. Time series prediction has always been one of the main application fields of machine learning algorithm. From the classical traditional time series prediction methods to long-term and short-term memory networks and closed-loop neural networks, the prediction ability is constantly improving. With the development of deep neural networks, convolution neural networks show superior performance in the prediction of time series. This paper proposes a new prediction model based on the improved WaveNet using not only the parameters of historical sales and hotel property, but also the parameters of holiday time and time position in the prediction range, which are processed by serialization. Simulation results are presented in details in this paper, where these results indicate the effectiveness of the proposed forecasting tool as an accurate technique.

In this paper, a new unbalanced decision tree algorithm for appealing expressions of reading poem is proposed. This algorithm called Weighted Division of Unbalanced Decision Tree (WDOUDT). Compared to the traditional decision tree, it has fewer nodes and more interpretability, faster convergence and higher accuracy. WDOUDT is used to identify the moving expression when students read poem. The results show that WDOUDT performed better in accuracy than decision tree, and its time complexity is lower than that of the classical decision tree algorithm. Besides, it has good generalization ability and good robustness against noise data.

With the development of society and economy, air conditioning has also become an integral part of our lives and it allows us to enjoy warmth in winter and cool in summer. However, there is also a phenomenon that people do not turn off the air conditioner due to personal savings awareness or anxious to go out. This paper intends to design a system where the air-conditioning room is automatically turned off if there is no one in half an hour. Pyroelectric sensor is used to detect the presence of people in the room, and wind speed sensor detects whether the air conditioner is on. If the room is empty and the air conditioner is not turned off for more than half an hour, the system is used to automatically turn off the air conditioner. It can reduce some waste of resources, it can also reduce certain environmental pollution, and realize economic and social benefits.

Modern service industry is one of the industries that widely applies artificial intelligent (AI) in the era of big data. The development of modern service industry plays a very important role in industrial transformation. The emergence of AI technology improves the effectiveness of data mining, and more and more scholars suggest a series of optimization methods to lift the prediction ability. In this study, tangent fruit fly optimization algorithm with step adjust (TANSAFOA) is proposed and it is used to optimize multivariate adaptive regression splines (MARS) and back propagation neural network (BPNN) to construct a prediction model of business performance. The result shows TANSAFOA can effectively optimize the prediction model and the BPNN model optimized by TANSAFOA has higher prediction performance than MARS. TANSAFOA BPNN-model is the most appropriate prediction model for modern service industry in China.

In order to break through the transmission capacity limitation of airborne computers, an airborne communication computer based on PowerPC+ SRIO Architecture is designed according to the wide application of the SRIO bus and the high performance of the PowerPC processor. In terms of optimization design, the computer adopts a redundant fault-tolerant system and improves the SRIO register initialization and address mapping method. The experimental results show that the transmission speed of the airborne communication computer based on PowerPC+ SRIO architecture reaches 81% of the theoretical value, and 16KBytes data can be transmitted in real time within 1μs.

According to the advantages of two-wheel independent drive electric vehicle that the torque and rotational speed of the drive wheel can be easily measured, a driving control strategy of driving antiskid control and torque coordination control is proposed based on road surface identification slip rate and adhesion coefficient. The principle of road surface identification is given. According to the signals of wheel speed sensor, lateral acceleration, longitudinal acceleration and torque sensor, the wheel slip ratio, road surface adhesion coefficient, vertical force and longitudinal force are calculated. Combined with μ-S function curve proposed by Burckhardt et al., the road surface adhesion coefficient and optimal slip ratio are identified. Driving antiskid control strategy adopts PID control technology based on road surface identification of optimal slip rate, peak adhesion coefficient and optimal slip rate, which mainly includes four parts: establishment of wheel stress model of two-wheel independently driven electric vehicle, approximate calculation of slip rate and road adhesion coefficient, optimal slip rate estimation based on fuzzy control and PID controller design based on optimal slip rate. The torque coordination control adopts PID control method based on ideal yaw rate.

At present, photovoltaic power generation mainly adopts a method of fixing the position and angle of the photovoltaic power generation panel and manually adjusting the angle of the photovoltaic power generation panel according to seasonal changes, but this power generation method has a relatively low working efficiency. Then came the sunflower-type photovoltaic power generation tracking system. Like the sunflower, this system can adjust its own angle by following the change of the sunlight angle at any time and place, so that the photovoltaic power generation panel always keeps running and working at an angle perpendicular to the sunlight . In order to further improve the efficiency of photovoltaic power generation and use solar energy more effectively, a sunflower-type photovoltaic power generation tracking system was designed and improved. Compared with the original system, in this system mode, the tracking accuracy of the photovoltaic power generation board can be further improved, and the photovoltaic power generation efficiency can be improved.

Surface defect detection is an important part in the process of aluminum production. The goal of a complete defect detection task is to realize the specific category and precise location of each defect in the image. Due to the similarity between surface defects and background, as well as the large diversity and difference in the appearance of the defects, this task is still challenging for applying this task in practice. In order to solve these problems, a deep learning based aluminum profile defect detection network is proposed. In order to realize strong classification ability, the deep convolutional neural network is used to generate feature graphs at each stage. On this basis, the variable feature pyramid module (DFP) is proposed, and deformable convolution is added to the feature layer to make the feature layer have the ability to adapt to defect deformation. Based on these multi-layer features, the size of anchor box was customized by Kmeans clustering algorithm, and then the region extraction network (RPN) was used to generate the region of interest (ROIs). For each ROI, a detector, consisting of a classifier and a bounding box regressor, produces the final detection results. Finally, we used a defect detection dataset of aluminum profiles to train and evaluate our approach. Through two sets of ablation experiments, the effectiveness of the introduced module is proved. Finally, we reached 76.92 mAP in aluminum profile data set.

In this paper, the main circuit topology of the three-phase four-wire active power filter (APF) and its harmonic current detection and control algorithm are studied. An active power filter system combining LCLCL filter and a three-phase H-bridge structure is proposed. This paper has applied harmonic detection algorithm based on an improved synchronous phase-locked loop (SRF-PLL) and recursive discrete Fourier transform (RDFT) while using triangular carrier modulation control based on PI adjustment to realize the harmonic compensation of the power filter to the grid current. The system has a notch effect on the harmonic at the switching frequency. It can detect the fundamental component and each harmonic component of the load current separately with a small amount of calculation, good real-time performance. Finally, a 5KVA single-phase APF prototype was built to test the superiority of the main circuit topology and the algorithm of current detection and control.

There are many classical application scenarios in computer vision, and human detection in dense scenarios is a classical problem. Its goal is to find all the people in an image or each frame of an video, as well as their positions and sizes, and frame them in a rectangular box. In recent years, deep learning has been applied to the field of computer vision, laying a solid foundation for human detection based on deep learning. Aiming at the crowdhuman dataset, based on faster-RCNN and feature pyramid network (FPN), this paper studies the training effect. In order to detect small objects, a multi-scale feature fusion network (MFFN) was proposed for feature extraction. According to the serious occlusion of dataset, a double-branch structure was proposed to improve the detection accuracy. These two structures combine to form a new network, we call it MDN, namely multi-feature fusion and double-branch network. Experiments show that these methods are helpful to improve the detection accuracy.

In order to reveal the influencing factors of the conflicts between the right-turning motor vehicles and the straight non-motor vehicles at the signalized intersection, the unmanned aerial vehicle (UAV) was used to collect the field data. And various methods such as manual statistics, video detection, deep learning were employed to analyze the data to obtain the traffic volume and the trajectory of pedestrians and vehicles. Pearson correlation analysis, 2-tailed significance test and other methods were used to examine the influences of traffic volume, speed, traffic control factors on traffic conflicts. The results show that (1) the correlation coefficient between the traffic volume of non-motor vehicles and the conflict frequency is 0.842 (close to 1), which indicates that the traffic flow of non-motor vehicle has a significant impact on the conflict frequency; (2) the speeds of right-turning vehicles at intersections have no significant impact on the conflict frequency (P = 0.071 > 0.05); (3) the setting of traffic markings is a key factor (P = 0.000 > 0.05), which can be seen in the traffic volume at intersections (not controlled by signals), and the improvement of traffic control elements can effectively reduce the frequency of conflict between motor and non-motor vehicles (MNV).

The simple network cable tester can only test one network cable at a time. When multiple network lines need to be tested, there are problems such as a long time, low accuracy, prool efficiency, and unable to display visually.A multi-network test circuit is developed and designed based on the principle of network cable testing on this paper. The principle of the simple network cable tester is briefly analyzed. The overall design idea, circuit design, and working principle of the LAN connectivity test circuit are explained in detail. The circuit characteristics are summarized. The actual application verification is performed during the management and maintenance of the unit LAN. The application results show that : this circuit has the characteristics of stable operation, simple operation, safety and reliability, convenient maintenance, and intuitive display. It improves the speed of network cable testing, increases the visual display of test results, improves the efficiency of network cable testing, and has achieved good application results.

Raft consensus algorithm is one of the commonly used consensus algorithms in distributed systems. It is mainly used to manage the consistency of log replication. It has the same function as Paxos, but compared to Paxos, Raft algorithm is easier to understand and easier to apply to actual systems. The Raft algorithm is a consensus algorithm adopted by the alliance chain. This article describes the details of Raft consensus algorithm and its application in detail.

With the rapid development of e-commerce, the "last mile" of logistics delivery has the problems of low efficiency and high cost. UVA delivery system can overcome this bottleneck whose system is a business process system which requires massive resources and real-time response. Edge-cloud computing integrates the massive processing power of cloud computing with the advantages of low latency and localization of edge computing. However, energy of UVAs is limited and UVAs are vulnerable to attacks. To address such issues, this paper proposes a secure model based on edge-cloud computing. Second, a new fitness function is proposed to evaluate the energy consumption and the security of task execution. Finally, a security-aware energy-efficient multi-source workflow task scheduling strategy within response time constraint is proposed. The results show that our strategy can achieve the security requirement and has less energy consumption under the limitation of response time.

Automobile noise has become the first source of environmental noise pollution, and its radiated power reaches 75% of the total environmental noise, while powertrain noise is the main source of automobile noise. With the problem of noise reduction of automobile powertrain, a 3D grid positioning technology, based on energy boundary element method, was proposed to quantify the noise source of automobile powertrain, and the generalized sound holography was applied to obtain the noise cloud map and analyze the noise propagation path, so as to determine the acoustic package material and the optimal location. Through the optimization experiment and simulation of 4 schemes, the acoustic package of automobile powertrain with the best effect was obtained, and the insertion loss of 6.7db could be obtained at 7m of the powertrain. Therefore, the results have important engineering application values for automobile noise reduction.

The process of intellectualization, information and refinemention of cement quality management in China develops slowly. This paper aims at this problem, Taking Zhonglian* Company as an example,on the basis of careful investigation and analysis of system functional requirements, based on the thought of modern quality management, it conducts an in-depth study of it, from the entry of cement raw materials to the finished cement products. The detailed design of the comprehensive quality management system laid the foundation for China's transition from a large cement country to a strong cement country.

Energy harvesting wireless sensor network (EH-WSN) uses a large number of sensor nodes to conduct environmental perception, data collection and data transmission through wireless links. However, the battery capacity of traditional energy harvesting wireless sensor is limited. In this paper, a kind of energy collection wireless sensor of dual-battery system is studied, and a centralized energy-aware routing algorithm based on reinforcement learning is proposed. Considering the energy of the overall link, each source node chooses a good path at the beginning of data transmission. After all the data is transmitted to the sink node, it selects the path according to the whole network state, and learns several times until it finds the best path choice. The simulation results show that the end-to-end delay of this protocol is lower than that of random packet forwarding.

This Study Focuses on the Composition, Performance, Application Field and Performance Simulation of Automotive & Automatic Navigation Control System, and Based on the Control Concept of the Specific Scheme, the Design Execution Leader is Used to Study the Redesign of the System Feedback. Finally, the Digital-Analog Simulation System is Applied to Improve the Control System, in Order to Provide a Feasible Reference for the Design and Performance and Improvement of the Modern Automobile Automatic Navigation Control System.

the Main Cause of Substation Circuit Breaker Break Accidents Mainly in Post Insulator Accidents is That the Flexural Strength of the Insulators Does Not Meet the Design Requirements. At the Same Time, There Are Problems Such as Eccentricity of the Porcelain Column, Large Difference in the Thickness of the Cement Casting, No Asphalt Buffer Layer in the Pouring Part, and No Waterproof Silicone Rubber. as the Temperature Changes, a Break Occurs under the Mechanical Force of the Switch. This Paper is Based on the Dynamic Simulation and Optimization of the Structure of High Voltage Circuit Breaker Based on the Finite Element Partial Dynamic Characteristics of Mechanical Structure. through Experiments by Ansys, the Optimized Solution Helps Manufacturers to Improve the Manufacturing Process of Insulators. the Operating Unit Strengthens the Equipment Acceptance, Operation, and Maintenance, and Carries out Visual Inspection and Ultrasonic Flaw Detection in Strict Accordance with the Requirements of the Regulations, Which Can Prevent the Occurrence of Insulator Breakage Accidents.

with the Rapid Development of Social Economy and the Continuous Improvement of People's Living Standards, People Pay More and More Attention to the Quality of Life, and Put Forward Higher Requirements for the Quality of Products, Especially the Quality of Cold Chain Products. as a Guarantee of the Quality of Cold Chain Products, Cold Chain Distribution Has Received the Attention of Enterprises. Based on the Technology of Internet of Things, This Paper Establishes a Path Optimization Model for Cold Chain Logistics by Adding Refrigeration Cost and Cargo Loss Cost to the Traditional Path Optimization Model. the Experimental Results Show That the Real-Time Route Optimization Method for Refrigerated Vehicles under the Internet of Things Environment is Effective. the Method Has Practical Significance in Achieving Fine Control of Cold Chain Logistics Distribution Costs and Improving Distribution Service Efficiency.

the Development of Design Has Been Accompanied by the Progress of Human Civilization and Culture. the Progress of Contemporary Society is Gradually Changing All Aspects of Human Production and Life. Living Ceramics is One of the Types of Ceramic Products, with Certain Practicality and Appreciation. It is a Relatively Traditional and Ancient Ceramic Product in Our Country. Life Ceramic Design Serves People, and Humanization Must Become the Core of All Designs. in the Contemporary Era, the Research on the Concept of Humanized Design Has Gone Deep into Many Fields and is Quite Mature. in Ceramic Modeling Design, the Attention of Humanized Design is Particularly Important. Virtual Reality (Vr) Technology Can Be Used as a Tool to Test the Performance of Ceramic Products, Which Can Be Evaluated and Corrected in All Aspects Before They Are Put into Production, So That Designers and Customers Can Feel and Identify the Advantages and Disadvantages of the Products through the Virtual Model on the Computer. in This Paper, Virtual Reality Technology is Applied in the Research of Life Ceramic Design, and the Application of Humanized Design in Life Ceramic Product Design is Analyzed.

with the Increasingly Fierce Competition in Science and Technology, the Competitiveness of Enterprises Depends on the Level of Products, and the Level of Products Depends on the Design of Products. Therefore, the Design Quality of Products Will Directly Affect Product Performance, Product Development Cycle and Economic Benefits, and Even the Development of National Economy. in Order to Achieve Product Innovation and Meet the Needs of Customers, This Paper Studies How to Integrate QFD and TRIZ to Achieve Mechanical Product Innovation Design. through the Research, It is Found That QFD Can Only Obtain Customer Demand, Not Create Demand, Only Find Conflict, Not Solve Conflict, and TRIZ Can Solve These Two Problems Well. Therefore, This Paper Takes HoQ as the Skeleton Framework, Integrates QFD and TRIZ in the Demand side of HoQ and the Roof to Guide the Product Design Process in Reality, Which Can Realize the Complementary Advantages, Overcome the Shortcomings of Only Relying on a Single Design Theory to Guide the Design Process, and Has Practical Significance.

Practical Teaching is Very Important to Vocational Education, But There Are Many Difficulties in Vocational Education Teaching. on the One Hand, It is Impossible for the School to Set Up Its Own Internship Factory for Its Own Major Due to the Restrictions of Capital Investment and Teaching Funds, Venues and Professionals. on the Other Hand, How to Find the Best Teaching Mode between the High Expectations of the Society for Vocational Education and the Students with Low General Scores?, Education Simulation Technology Can Meet the Requirements, the Simulation Technology in Vocational Education, the Author of This Paper How the Product Development, Design, Development Method and the Application of Simulation Technology in the Vocational Education Mode, Application Characteristic and Using Effect Were Discussed.

in the Process of Computer Network Security Maintenance, Firewall Technology is the Main Security Maintenance Approach and Can Play an Effective Protection Role. Due to the Popularity and Widespread Use of Computer Technology, Network Security Has Received Widespread Attention from All Walks of Life. the Purpose of This Article is to Study Firewall Technology in Computer Network Security Based on 5G Environment. 5G Mobile Networks Also Have Some Shortcomings, Such as Weak Network Penetration Capabilities. Once the Base Station is Not Fully Deployed, There Will Be Gaps in the Borders of Communication Cells, Resulting in Network Data Transmission Interruption. in Order to Solve the Above Problems, This Paper Proposes Mobile Network Optimization Measures. Based on Many Years of Practice and Exploration, It Focuses on Describing the Complex Communication Application Scenarios in Cities, and Introduces a Variety of Mobile Network Optimization Technologies to Achieve Seamless Network Coverage and Ensure That People Can Use Mobile Communication Networks Normally. the High-Speed, Large-Capacity, Low-Latency, and High-Reliability Services Brought by 5G Will Have an Impact on the Local Optical Cable Architecture and Core Utilization. According to the Experimental Results, the Effective Application of Firewall Technology Can Greatly Improve the Security of Computer Networks and Ensure the Security of Computer Networks to a Great Extent.

the Extensible Mark-Up Language Xml is a Simple Data Storage Language, Using a Series of Simple Tags to Describe Data, and These Markers Can Be Used to Establish a Convenient Way, Although the Xml Space to Occupy More Space Than the Binary Data, But Xml is Extremely Simple and Easy to Master and Use. Xml Organized Management of Web Information Resources Distributed in Logic, Web Information Retrieval System Structure to Design a Professional Content, and Provides a Way for Users to Obtain the Special Contents Oriented Information Resources. a Good Data Storage Format, Scalable, Highly Structured and Easy for Network Transmission Characteristics So, and Xml is Determined by the Excellent Performance. the Paper Presents Xml and Its Application in Library and Information Retrieval.

with the Rapid Development and Popularization of Computer and Network Technology, Today's Society is Developing in an All-Round Way. in Today's Society, the Amount of Information is Increasing Rapidly in Several Stages Every Day. in This Information Explosion Society, Information Automation Management Has Become the Trend of Today's Society. Books as an Important Carrier of Information, with the Development of Information Technology, the Types and Quantity of Books Have Also Changed Greatly, and Growing At a Very Fast Speed. How to Find, Borrow and Return Books More Efficiently and Quickly Has Become a Difficult Problem in Book Management. the Popularization of Computer and the Development of Network Technology Have Brought an Opportunity for This Problem. This Paper Analyzes the Design of the Management System Based on the Physical Structure of B/s and Introduces in Detail the Development Tools, System Function Requirements and System Software and Hardware Requirements for the Realization of the Web Management System of Children's Books. the System Has Been Realized by Developing Technology and Tools, Which Ensures the Normal Operation of the System, Realizes the Integrated Function of Information Service of the Library, Increases the Working Efficiency of the Management Personnel, Facilitates the Reading Demand of the Majority of Children, Reduces the Maintenance Cost of the Library, and Makes the Library Develop Towards the Modernization of Information Technology.

the Construction of Smart Grid Puts Forward Higher Requirements for the Safe Operation of Substation, But There Are Still Some Hidden Dangers of Human Factors in the Current Operation of Substation. the Safety Management System of Substation Operation Based on Wireless Positioning Technology Can Avoid the Influence of Human Factors by Technical Means. Based on the Analysis of the Special Requirements of Operation Safety Management in Substation Site, It is Pointed out That Css Broadband Wireless Positioning Technology is Suitable for Substation Application Because of Its High Positioning Accuracy, Strong Anti-Interference Ability and Reasonable Price. the Work Flow and Application of the Substation Operation Safety Management System Are Described.

The rapid economic and social development cannot be separated from the important contribution and guarantee of the power industry. In the development of power industry, substation technology plays a very important role. The development of in-depth learning based on mobile learning provides an effective means for data value extraction. Therefore, it is of great significance to guide power enterprises to make correct decisions by making full use of these data based on the actual power grid and analyzing them in depth to discover and utilize the hidden information. Firstly, the basic framework of in-depth learning is elaborated, and the theoretical basis and technical system of mobile learning are summarized. According to the actual demand of power system, this paper summarizes the key application fields and values of mobile learning-based deep learning in power system data processing based on image data and spatio-temporal data, and puts forward relevant technical development suggestions.

The computer simulation technology can reproduce the crowd movement trend under the emergency situation in a real all-round way, summarize the movement law, and formulate the emergency plan of public places. It can conduct scientific analysis and guidance on crowd evacuation in emergency situations through comprehensive evaluation of crowd evacuation channels in buildings and public places in combination with the emergency situations that can occur in the movement behavior simulation scenarios in emergency situations. In terms of security research hotspot, the emotion of personnel in emergency has a direct impact on crowd evacuation, so it is necessary to summarize Arnold's emotion model. The occurrence of adverse events in emergency should be avoided to achieve more efficient evacuation of personnel.

In order to support the intelligent combat capability of future wars, the network information system needs in-depth research and exploration in the field of artificial intelligence. This paper designs a distributed aggregation storage mode with the characteristics of balanced storage load distribution and local node aggregation storage under Big Table model. A distributed parallel query engine that uses the Group-By mode to distribute parallel computation query trees. A knowledge map completion method based on Bayesian inference is proposed. Bayesian probabilistic inference theory and RDF implication inference rules are used to jointly infer the potential relationships between entity nodes to predict the relationship between new nodes and original nodes, thus improving the mining efficiency of potential factors in the model and the accuracy of prediction of unknown relationships. Based on Big Table model, the entity sets stored row by row are evenly divided and stored through random prefix and pre-partition operations to realize load balancing. At the same time, random prefixes can be uniformly distributed to nodes for storing entities of the same type, and aggregated by entity category on a single node. With the continuous development of knowledge map technology, future knowledge map learning and reasoning technology can be integrated with new fields such as machine depth learning, cloud computing, block chain, big data, biological genetic engineering, etc. to play an important social value.

Based on the signal characteristics of the existing weighted fractional fourier transform (WFRFT) communication, this paper studies the equivalent signal to noise ratio (SNR) and bit error rate (BER) of the non-target receiver under the condition of the transformation parameter errors, with the purpose of measuring the anti-parameter scanning characteristics of the single-parameter WFRFT signal. The constellation splitting characteristics of multi-parameter WFRFT signal are further studied.Multi-parameter WFRFT modulation can not only improve the resistance of single-parameter system to parameter scanning detection, but also effectively hide the original modulation mode due to its characteristic of splitting modulation constellation.

As a kind of vertical transportation device, elevator is widely used in the production field and daily lives. In order to detect the reset time of the elevator hydraulic buffer, this article first analyzes the shortcomings of the existing detection methods, then provides a design of detection system for the elevator hydraulic buffer based on intelligent image recognition. This system can detect the reset time and reset distance, and grasp the dynamic compression and reposition process of the elevator hydraulic buffer. Besides, this system can also bring lots of advantages, such as wide universality, easy installation, high degree intelligence, less influence of human factors and no risk for the inspectors in measurement process.

Due to the interference of the external environment such as rainy weather on the camera, raindrops can easily adhere to the lens and seriously affect the quality of the photos taken. Therefore, it is of great significance to remove raindrops from the image and improve the quality of the photo. In this paper, a raindrop method for generative adversarial network images based on differential learning is proposed. The general generative network is to input images with raindrops and output clean images. The generative network in this paper does not directly output clean images, but learning the difference between images with raindrops and without raindrops, then subtract the learned difference from the image with raindrops to generate a clean image. In order to learn this difference more effectively, adding reconstruction loss to the generative network, the pre-trained VGG-16 network is used to extract the difference between the generated image and the real image features and calculate the mean square error. The experimental results show, the method in this paper can not only remove the raindrops in the image well, but also reconstruct the image information of the part blocked by the raindrops. The image processed by the algorithm in this paper is tested using the yolov3 target detection algorithm, which can significantly improve the recognition accuracy of the detection algorithm.

Aiming at moving object detection in Airborne Hyperspectral Remote Sensing Images under complex background, this paper proposes a background removal method based on Mixture Gauss Model. Firstly, in order to eliminate the influence of the dithering of airborne platform on remote sensing image, this paper uses affine transform-based image registration algorithm to achieve the image registration of 3D data. Then, this paper selects specific spectral images, establishes the Mixture Gauss background model, and subtracts the background to achieve the extraction of moving objects. The experimental results show that the method can extract moving objects in Airborne Hyperspectral Remote Sensing Images effectively.

In this paper, we propose a road segmentation model using deep learning technology. The model is essentially a U-net with pre-trained DenseNet-169 encoder. Multi-scale and high-level semantic information are extracted effectively by dense residual learning and attention mechanism. In addition, an improved focal loss function is proposed to handle extremely imbalanced road samples. Experimental results demonstrate that our proposed road segmentation model can accurately extract complex road areas in remote sensing images and has IoU of 0.6308 which ranks 22st in the DeepGlobe Road Extraction Challenge leaderboard at the time of writing.

Aiming at parameter estimation of LFM signal in low SNR, a new method based on short time Fourier transform (STFT) and random sample consensus (RANSAC) is proposed in this paper. STFT is carried out to extract time-frequency curve firstly. Then median filter is used to eliminate glitches and disturbs. Finally, RANSAC process is applied to construct the model of the chirp line and obtain the parameters. In the proposal, the use of median filter and RANSAC iterations improve the adaptability to extremely low SNR. Simulation results confirmed the effectiveness of the proposed method.

Long-term monitoring of fetal health status is important for high-risk pregnancies. In order to extract the pure fetal phonocardiogram(fPCG) and increase the calculating accuracy of fetal heart rate (FHR), a combination of Empirical Mode Decomposition (EMD), Cepstrum, Lifting Wavelet (LWT) and Hilbert transform (HT) method was proposed. First, the heart sound signal was decomposed into finite intrinsic mode functions (IMF) by EMD, and then the signal de-noising was implemented by LWT. Second, the envelope of the signal was obtained by HHT. Finally, the fetal heart rate was obtained by cepstrum analysis. The simulation result shows that the fetal heart rate extracted by this integrated signal processing method is right, and it enhances the precision of fetal heart rate greatly. The wavelet de-noising technology can reduce the noise very well with small computation. It is concluded that this integrated signal method is capable of extracting feature from fPCG.

A deep learning model has a large number of free parameters, which need to be effectively trained on a large number of samples to calculate the depth parameters. However, many special applications like underwater acoustic signal recongnition cannot provide enough dataset to guarantee high performance. In addition, the original dataset adopts some formats, such as audio, which makes it difficult to capture features. To overcome these challenges, we propose a novel framework. Firstly, based on the evaluation of spectrum method, our framework selects the appropriate preprocessing method. Then it modifies GAN to generate samples, and establishes an independent classification network to ensure the quality of samples. Finally, our framework applies the existing classification network to evaluate performance and selects the best one for the LOFAR spectrum. The experimental results demonstrate that the proposed method can generate high-quality LOFAR spectrum and improve the prediction accuracy of the classification model significantly.

In recent years, automatic detection technology of motor vehicles has developed rapidly. However, due to some interference problems, there is little research on the automatic detection technology of non-motor vehicle. This paper proposes a method which based on the experience and results of actual projects for discriminating non-motor vehicles in real-time video, detecting and recognizing license plates. The algorithm and steps will be described in detail. The image difference method and fractional differential method are combined to extract the contours of dynamic non-motor vehicles in the video, and the methods are used to filter the contours of non-motor vehicles using skin color filter detection and geometric feature discrimination to assist the training of cascade neural networks. Non-motor vehicle license plate image clustering is detected in the extracted contour by clustering the background color of the license plate and determining the ratio of the rectangle surrounding the license plate. The boundary expansion method is used in combination with the Faster R-CNN (Faster-Convolutional Neural Networks) to train the model, and then BP (Back propagation) neural network is used to identify characters in the target area. And a trajectory tracking method is proposed to automatically determine whether non-motorized vehicles have violations.

This paper studies the structure and characteristics of LANDSAT8 OLI data, and transfers Deep Belief Networks model to OLI datasets. Firstly, the sub-sampling window is used to filter the pretraining data, which reduces the calculation cost of the hidden layer feature representation of the model. The parameters of each layer of DBN are trained forward, optimization parameters are used to make the error converge to a low value in back propagation. Finally, the activation function is used to reduce the dimension of the hidden layer of the model, and the output vector is used as the global feature representation of the image. Running DBN to complete the classification of LANDSAT8 remote sensing image in the study area, the experimental results show that the model has completed the classification and achieved high user accuracy and reliability in the case of that all types of the wetlands have similar spectral characteristics. The conclusion is that the pretrained features can be well applied to the LANDSAT8 remote sensing datasets. The accuracy is improved within the scope of cost calculation, and the local minina problem will also occur in deepening model.

The system is designed to construct a Function Signal Generator based on Direct Digital Frequency Synthesis Technology (DDFS), in which MCU and FPGA are the key components and necessary peripheral circuits are provided. Its main modules are responsible for the formation of oscillogram, the frequency synthesis, the digital-to-analog transformation, the control of range, and back-stage treatment, the user interface and other necessary modules. MCU controls FPGA to generate the oscillogram and other control words, and the FPGA controls output waveform and frequency synthesis. Amplitude control is realized by 2 DA. Low-pass filter and Transistor push-pull power amplification circuit are used to increase the quality of the output waveform and its loading capacity, and display the signal on the LCD. The system constitutes of 2 DAC and the FPGA, so that realizes the expanded requirements of 1 Hz-1MHz of the four waveform output.

Sleep apnea syndrome is a serious threat to the life and health of patients. In this paper, we designed a sleep apnea monitoring system based on flexible fabric electrode, and completed the specific hardware design including flexible fabric electrode, ECG signal conditioning module, microcontroller module and Bluetooth module. Among them, ADS1292 is mainly used for ECG signal conditioning, and EFM32 is selected as microcontroller. Finally, the experimental results show that the collected signal can clearly identify the ECG characteristics such as QRS wave, and meet the requirements of the subsequent research of ECG signal. The terminal equipment we designed is low cost, comfortable to wear, and can be monitored for a long time, which is suitable for daily household monitoring.

The automatic classification of ECG images has great significance for doctors to diagnose cardiac diseases. In order to improve the accuracy and efficiency of disease diagnosis, this paper presents an automatic classification system for ECG arrhythmia images. The database of MIT-BIH is processed visually and a waveform detection method is proposed for detecting the QRS waveform. Wavelet transform is used to detect the R-R interval and locate the R point peak. A convolutional neural network(CNN) model was built to train and classify the ECG images. Experimental results show that according to the ANSI/AAMI EC57 evaluation criteria, the accuracy(Acc) rate of ventricular ectopic beat(VEB) can reach 95.9%. The sensitivity(Se) evaluation index is 93.0%. The specificity(Spe) evaluation index is 91.9%. For the supraventricular ectopic beat(SVEB) class, the accuracy rate is 93.2%, the sensitivity evaluation index is 81.3%, and the specificity evaluation index is 90.5%.

Bioacoustics has recently become one of the "big data" research topics since many bird monitoring projects have collected terabytes of audio using remote sensors. The challenge in recent years is to develop algorithms to realize fully automatic recognition of bird species through analysing environmental recordings. A number of approaches directly draw on experience of effective algorithms in signal processing and image processing areas. They seem working well for small data or lab data, however, the outcomes for large-scale environmental data shows a big gap between theoretical experiments and real applications. To provide possible clues for future research, we review the state-of-art development in automated bird species recognition, and identify wide range of algorithms on noise removal, bird call detection, feature extraction for classification. The significant software tools and publicly available datasets for the task are presented. This survey can be valuable for new researchers who are about to start the journey with birdsong analysis.

This audio signal analyzer is based on a microcontroller (MCU) and a field programmable gate array (FPGA). It consists of a controllable gain amplifier circuit, an active filter circuit, A / D conversion and D / A feedback circuits, and a main control circuit and liquid crystal display module. Through AD conversion, the audio signal is sampled, the continuous signal is discretized, and then the FFT fast Fourier transform operation is performed to analyze and process each frequency component and power of the audio signal in the time and frequency domain. By controlling the gain of the amplifier to expand the dynamic range of the input signal and improve the sensitivity, it can also complete the corresponding human-computer interaction control based on the PS2 keyboard input, while using the high-resolution LCD to display the signal spectrum. The audio signal frequency range that this system can accurately measure is 20Hz-10KHz, its amplitude range is 5mVpp-5Vpp, and the resolution is divided into 20Hz and 100Hz. The measured power accuracy is as high as 1%, and the period of the periodic signal can be accurately measured, which is the ideal solution for audio signal analyzers.

Embedded AI rised in recent years, AI algorithm deployment platforms need strong computing power, but embedded systems focus on the balance between performance and cost. The contradiction has become a great challenge for the development of embedded AI. This paper analyzes the "ARM+FPGA" architecture and NEON register resources of the ZYNQ series chips introduced by Xilinx, and proposes a method to reasonably schedule the above resources to accelerate video preprocessing when it is used as an edge processing platform for intelligent monitoring terminals. This method makes the video intelligent analysis process of the monitoring terminal achieve real-time.

In the process of Augmented Reality (AR) image registration, the traditional ORB (oriented FAST and rotated BRIEF) algorithm has low registration rate and poor real-time performance. In this paper, an improved AR image registration method based on improved ORB is proposed. Firstly, the calibration image and video frame image feature points are obtained by the improved FAST feature detection algorithm. Then, the binary descriptor of BRISK, which using the custom domain sampling pattern is used for feature description, and the scale invariance of the traditional ORB algorithm is improved. Finally, the random sampling consistency (RANSAC) algorithm is used to eliminate the wrong matching point pairs and optimize the feature matching. Experiments show that compared with the AR image registration method described by the traditional ORB algorithm and the FREAK feature, the registration rate of the proposed algorithm is increased by 1.1% and 8.4%, and the generation time is reduced by 0.13s and 0.12s, respectively. The experimental results show that the AR image registration method proposed in this paper can obtain higher feature point registration rate, and has better real-time performance, which can better meet the application needs of AR image registration.

In order to describe uncertain and fuzzy emotions in recognizing facial expressions, a PAD regressive model based on deep convolutional neural network model is built to quantize the emotions, by which the facial expressions can be mapped to the emotional space of PAD. Then the emotional membership function is proposed to describe the uncertainty and fuzziness of the emotions in the space of PAD. Finally the testing results demonstrate that the regressive model performs well in recognizing facial expressions and the results from describing the emotions quantitatively agree with the human cognition to facial expressions.

Web image character recognition has important application value in the commercial field and text detection is the basis of extracting this text information. Among many text detection algorithms, EAST (efficient and accurate scene text detector) algorithm has a good performance in natural scenes, but it is ordinary in web pictures. Based on this, aiming at the problem that the number of channels in the QUAD output layer of EAST algorithm is redundant, which leads to the low accuracy of web image text detection, it is modified as the distance from the point marked as a positive sample to the four vertices of the text box where it is located instead of the coordinate offset to the vertices. The improvement reduces from the original 8 channels to 4 channels, which restricts the optimization direction of the model, and then by setting a new loss function in order to improve the loss value of the edge area of the text box and the pixels in the image that are difficult to detect, the model is more suitable for the text detection of the web image. The experimental results show that the improved EAST algorithm significantly improves the accuracy of text detection in web images.

Diagnosis and recognition of rice diseases and pests, amongst others, are an essential research direction of precision agriculture, which help to effectively prevent and control diseases and pests as well as improve rice yield and quality. Starting with the basic principles and technical routes of feature extraction and classification recognition of rice disease images, this paper expounds characteristics, realization methods and applications of many current feature extraction and recognition methods in China and abroad, analyzes the advantages of and problems in applications of different methods and technologies into recognition of rice diseases and pests, analyzes problems to be solved, proposes corresponding solutions, and looks into the development future of this research.

The insulator fault of high voltage line is the main factor of transmission accidents, so positioning and detection of burst insulator of power line has become an important part of routine detection. Traditional UAV detection is still mainly to evaluate the image transmitted by UAV manually, which is not only time-consuming, but also not accurate. In this paper, UAV aerial images are used to make datasets, the problem of insulator class imbalance (normal insulator and burst insulator) in the model training process is solved by using data augmentation, an improved depth learning algorithm is proposed based on YOLOv3, which provides rich semantic information for prediction layer by adding feature mapping module. At the same time, the residual network is introduced into the feature extraction, which improves the detection accuracy of small objects, more effectively extracts the object features of burst insulator, carries out real-time object detection and positioning, and completes the daily detection of insulator state of power transmission lines. The experimental results show that the detection accuracy (mAP) of the improved YOLOv3 algorithm is 91.22%, and the detection speed is 28 frames/s. The recall rate and Intersection Over Union (IOU) are also improved.

As to the problem of surface reconstruction in 3D laser scanner, we propose a divide-and-conquer based method, and realize the continuing growth of the reconstructed surface through comprehensive analysis of the correlations of scanning line point cloud data. First of all, filter the scanning line point cloud data, then confirm the relationship between the newly-insert point and the generated surface by using the local search algorithm based on adaptive point cloud segmentation, and finally, classify the newly-inserted points as sparse points, density points, growth points and flyback points by using the given maximum and minimum distance thresholds, and then the specific surface reconstruction is performed according to the point's classification. The experimental results show that the local search algorithm greatly reduces the calculation of the insert location of the point cloud compared with the global traversal search algorithm; the proposed method can process the flyback points, enhancing the repairability and the expandability of the reconstructed 3D surface compared with the traditional method. Furthermore, the quality of the reconstructed 3D surface, the generating speed and the occupation of storage space can all be controlled by adjusting the maximum and minimum distance thresholds according to actual demand.

With the Continuous Improvement of Medical Information, the Continuous Development of Image Processing Technology and the Continuous Upgrading of Hospital Image Acquisition Equipment, Medical Image Detection Has Become More and More Reliable and Effective, Becoming an Important Basis for Doctors' Diagnosis. Now Large Hospitals Use the Combination of Radiology Information Management System and Image Archiving and Communication System to Store and Access Patients' Images. with the Establishment of Regional Image Sharing and Cooperation Platform, the Online - Near Line - Offline Storage Strategy Adopted by the Traditional Image Archiving and Communication System Can No Longer Meet the Needs of Mass Image Data Storage. in the Retrieval of Mass Image Data, the Relational Database Used by the Image Archiving and Communication System is Inefficient in Reading the Image on the Scale of Mass Data, and the Existing Image the Archive and Communication System Lacks Complete Data Backup and Disaster Recovery Scheme. This Paper Analyses the Feasibility, Advantages and Problems to Be Solved of Using Hdfs to Store Massive Medical Image Files, and Studies How to Realize the Function of Precise Access and Range Access Provided by Pacs Efficiently in Hadoop.

Pedestrian Detection is a sub-task in the object detection due to it plays an important role in video surveillance. This project train model algorithm by using offline retail data and intelligent identification based on the offline retail brand store scene and focuses on the two core personnel of the retail scene-shopping guides and customers. This project detects pedestrian's localization and their key points by using deep learning and computer vision technologies.

In view of the fact that the existing image quality evaluation methods are generally artificial design features, it is difficult to automatically and effectively extract image features that conform to the human visual system. Inspired by human visual characteristics, a new full reference image quality evaluation method based on depth learning model based on core concepts is proposed. Firstly, depth learning algorithm is used to extract multi-layer features from reference images and distorted images respectively. Then, the local similarity of the feature map of the reference image and the distorted image in each layer is calculated as the local quality description of the corresponding depth. Finally, the local quality of all layers is synthesized to obtain the overall quality score of the image. On the basis of the pre-training model, the depth model network is fine-tuned by using the image visual evaluation data set to obtain a depth model for evaluation. The standard experiment shows that fine-tuning training of each pre-training model on the standard data set achieves good classification results. Experiments show that the designed depth learning model based on core concepts is superior to the existing full reference image quality evaluation methods, and its prediction results have good accuracy and consistency with subjective quality evaluation.

This paper proposes a HBase connection dynamic keeping method based on the Reactor mode. This method uses the Reactor mode to implement the client connection maintenance mechanism based on the NSQ message queue. By only creating the TCP control block of the client connection and buffering it to the NSQ queue, it improves the response of the user connection and the concurrent processing capacity. By monitoring the active state of the connected TCP block in the message queue, the real-time processing capability of the connection is guaranteed. The reuse of HBase connection objects is realized through the listen-back callback mechanism, and the CPU and memory resource utilization is improved by dynamically adjusting the size of the connection object pool.

With the spread of new coronavirus, new forms of online teaching are widely used in colleges and universities in China. Teachers are the main parts relying on the existing online teaching platform, which can make up for the defect of single online platform, simulate off-line teaching better, complement the advantage of off-line teaching, and improve the quality of online teaching effectively by using the new hybrid teaching mode of program language courses with other network tools software.

Object detection in remote sensing image is a challenging task in computer vision. Remote sensing images have different characteristics compared with conventional images. Especially object detection in remote sensing image needs to focus on small targets with different ratio and orientation. In this paper, we propose a novel detection architecture for remote sensing imagery to solve the problem of scale diversity. On the basis of Cascade R-CNN, we have developed Parallel Cascade R-CNN. In the second stage, parallel detection heads are used for separate detection, and their RoIAlign modules have different output sizes. In addition, different preprocessing methods are applied according to the shape and quantity characteristics of different classes. We evaluated our algorithm on DOTA dataset. Experiments have shown that our algorithm can achieve performance improvement on a high baseline. And the detection performance of those categories with smaller objects has been improved. In the detection task of horizontal bounding box, we obtained the mAP of 78.96%, which reached the state of the art. Our algorithm is simple and has good performance, easy to be migrated to various network structures.

Cognitive radio (CR) is a hopeful technology to reduce the negative effect of spectrum scarcity caused by the enormous amount of wireless mobile device. Spectrum sensing could alleviate the interference and harm for primary user, and enhance the wireless access capability. This paper proposes a novel spectrum sensing method by using the summation of multiple large eigenvalues (SMLE) and stochastic resonance (SR). SR is used to enforce the detection signal of multiple antennas in low SNR condition. Then the sample covariance matrix of the enforcing detection signals and its multiple large eigenvalues are computed for constructing test statistic. The simulation results demonstrate that the proposed detector based on SMLE and SR is superior than the existing detector based on SMLE, and is robust in strong noise background.

Ultrasound imaging guidance has been widely used in peripheral nerve block to guide the anaesthetist to insert the needle and observe injection fluid diffusion. However, needle-nerve contact(NNC) positions visualization is difficult to achieve because of requirement of high image solution and expert. To better guide the anaesthetist to predict the distance from needle to nerve and reduce the needle trauma, we attempted to use Needle tip force(NTF) and Opening injection pressure (OIP) to identify the NNC positions under ultrasound guidance. The relationship between OIP and NNC, NTF and NNC, NTF and OIP is unclear now, and we conducted a prospective, observational study to define the relationship. In this study, we designed a multi-function block needle and built a real-time NTF and OIP measurement which could trace and record NTF and OIP data in the simulator sciatic nerve. NNC positions were divided into three distinct NNC positions: Pre-NNC, First-NNC, Forceful-NNC using ultrasound guidance image. We statistically analysis NTF, OIP in the three NNC positions, and the results indicate the NTF differentiate in three NNC positions(p<0.001), OIP differentiate as well (p<0.001) and the NTF has a highly positive with the OIP in the same NNC positions with Pearson correlation 0.62.

Gesture recognition based on visual modal often encounters the problem of reduced recognition rate in some extreme environments such as in a dim or near-skinned background. When human beings make judgments, they will integrate various modal information. There should also be some connections between human gestures and speech. Based on this, we propose a multimodal gesture recognition network. We use 3D CNN to extract visual features, GRU to extract speech features, and fuse them at late stage to make the final judgment. At the same time, we use a two-stage structure, a shallow network as detector and a deep network as classifier to reduce the memory usage and energy consumption. We make a gesture dataset recorded in a dim environment, named DarkGesture. In this dataset, people say the gesture's name when they make a gesture. Then, the network proposed in this paper is compared with the single-modal recognition network based on DarkGesture. The results show that the multi-modal recognition network proposed in this paper has better recognition effect.

The accurate recognition of insects is of great significance for the protection and treatment of crops, vegetables and fruit trees. In order to improve the accuracy of insect images classification at the order level, the insect recognition algorithm based on complex frequency domain transform and multi-feature fusion is proposed in this paper. Firstly, the color features and Histogram of Oriented Gradient features of the insect images were extracted in the spatial domain. Secondly, the multi-scale and multi-directional complex frequency domain transform was performed on the images using the Multi-directional Dual Tree Complex Wavelet Transform. The shape features and texture features were obtained from the low frequency sub-band images. Finally, the spatial domain features and the complex frequency domain features were fused using the Relief-F algorithm, and the Support Vector Machine was used to recognition the insect images at the order level. The results of insect images recognition at nine orders level showed that the proposed method of multi-feature fusion in complex frequency domain and spatial domain can significantly improve the accuracy of insect recognition.

Traditional signal detection methods have achieved satisfactory performance in many contexts of signal processing. However, these methods based on mathematical statistics show a drawback in dealing with the low SNR cases, which limits their practicability. To this end, inspired by image processing techniques, we first make use of the short time Fourier transform to generate sufficient 2-D spectrograms of the received data. Then we extract high dimensional features of these spectrogram to construct high dimensional covariance matrices, transforming into a binary classification problem lying on a symmetric positive definite (SPD) manifold. In addition, by reducing the dimensionality directly on the SPD manifold, these spectrograms are mapped into a more discriminative SPD manifold, which improves the separability between the two classes. The simulation experiment results demonstrate that our method achieve satisfactory signal detection performance in the task of signal detection under K distribution data, even in the case of SNR = -10.

Aiming at the reconnaissance of ultra-widaband (UWB) LFM signal, a method using dual-channel joint working mode is introduced and a parameter estimation algorithm of cross-channel signal is proposed in this paper. Short time Fourier transform (STFT) is carried out to extract time-frequency curve firstly. Then least-square (LS) algorithm is used to roughly estimate the parameters of signal in single channel and whether it is a cross-channel signal is determined. Finally, random sample consensus (RANSAC) iteration is applied to construct the model of the chirp line and obtain the accurate parameters. Simulation results confirmed the effectiveness of the proposed method.

With the continuous development of social networks and the continuous emergence of public opinion on the Internet, the government's emergency response is facing tremendous pressure. In this paper, cellular automata is used to compare the social relationships between individuals in the network. Combustion propagation models are used to compare the network masses to several types of burning substances according to their propagation characteristics. The influencing factors among individuals are determined. A network public opinion propagation model is established to explore the rationality of its propagation.

In order to improve the detection efficiency and accuracy of business logic vulnerability, a test method based on behavior characteristics was proposed. It uses the test accountes to crawl url of business system, analyzes the characteristics of request sequence, parameter attributes, request parameters, response parameters, relationship between parameters, and digs out the vulnerability, and then uses the test elements to verify whether there is a vulnerability. Experimental and analytical results show that this method which is instead of manual method can effectively detect the vulnerability of invalid identity authentication, invalid access control and sensitive information leakage.

Diabetic retinopathy, a complication of diabetes, is a significant cause of vision loss and blindness. Early detection of diabetic retinopathy can help reduce the risk of blinding. However, automatic diabetic retinopathy identification is a challenging task due to their different morphology during a different stage. Aiming at the problem of low efficiency for most of the existing methods, we developed a diabetic retinopathy recognition system based on transfer learning. The system utilizes transfer learning, which trains a neural network based on the DenseNet201 network model and Messidor Data Set, and can not only train an active network quickly, but also has a reasonable effect on the classification of diabetic retinopathy.

Environment perception based on vision plays an important role in autonomous driving technology. Although vision perception has achieved certain results in recent years, many methods can not solve the contradiction between speed and precision. In this paper, we propose a system for fast and accurate object detection and localization based on binocular vision. For object detection, a neural network model based on YOLOv3 is proposed. Specifically, MobileNet is employed in the backbone of YOLOv3 to improve the speed of feature extraction. Then the corresponding ORB feature points are extracted from continuous stereo images which take from the binocular cameras on the moving car. Thus, the disparity of each ORB feature point is calculated. After that, we use the result of object detection to screen the ORB feature points. Finally, the depth of the targets in the traffic scene can be estimated. Experiments on the KITTI dataset show the efficacy of our system, as well as the accuracy and robustness of our object localization relative to ground truth and prior works.

Flexible eddy current array nondestructive testing technology has the advantages of large single scanning area and fast scanning speed in the defect detection of conductive materials, so it becomes a common research hotspot of sensor and nondestructive testing technology. The performance of the array probe on the sensor is an important factor to determine the performance of the sensor, and the detection circuit and excitation mode are another important factor. In this paper, the array coils with different parameters are simulated and analyzed, and the resonant detection circuit based on ldc1614c digital inductance conversion chip is used to detect the high permeability materials on curved surface. Through the digital imaging technology, the defect image of the tested material is obtained to verify the selection of the probe coil and the reliability of the detection circuit.

Scene parsing and robotic manipulation using visual sensing are of significant importance for mobile and robotics related applications. And increasing efforts are being made to associate semantic concepts to geometric entities in robot's surroundings. However, they are not well suitable for resource-constrained and mobile platforms. Toward an efficient semantic mapping solution, we propose a specifically light-weight segmentation network for building a dense 3D semantic map in real-time. The propose architecture is based on depth-wise convolution and channel shuffle to reduce computation cost. An improved Atrous Spatial Pyramid Pooling (ASPP) and encoder-decoder structures are presented and multi-scale features are merged for better prediction. Experiments are compared with other methods to prove the effectiveness and superiority.

Cuffless method for blood pressure measurement is an important methods for continuous health status monitoring. A pulse wave is a periodic time-series signal that reflects a non-linear, non-stationary change in the pulse signal over time. Traditional ways of pulse wave based blood pressure assessment rely on feature extraction from pulse signals, which are usually signal quality dependent and lack of consistence among studies. In this paper, a method of blood pressure measurement of using continuous pulse waveform and long-term and short-term memory network is proposed, which avoids the process of manually extracting waveform features. Experiments were performed with both pulse wave signals and the arterial blood pressure signals form the MIMIC database. Empirical mode decomposition was applied for signal preprocessing, and the time series of the pulse wave was analyzed to establish a Long Short-Term Memory neural network for blood pressure assessment. An average prediction accuracy of 83.2% was achieved.

Convolutional Neural Networks (CNN) have achieved great performance in many visual tasks. However, CNN models are sensitive to samples with large spatial variants, especially severe in fine-grained classification task. In this paper, we propose a novel CNN model called ST-BCNN to solve these problems. ST-BCNN contains two functional CNN modules: Spatial Transform Network (STN) and Bilinear CNN(BCNN). Firstly, STN module is used to select key region in input samples and get it spatially modified. Since the adoption of STN will cause an information loss phenomenon called boundary loss, we design a brand-new IOU loss method to solve it. We make a theoretical analysis of the IOU loss method. Secondly, to discover discriminative features for fine-grained classification task, BCNN module is applied. BCNN interacts CNN features from different channels to produce more discriminative bilinear features than fully connected features of CNN. ST-BCNN works by reducing irrelevant spatial states and producing fine-grained features. We evaluate our model on 3 public fine-grained classification datasets with large spatial variants: CUB200-2011, Fish100 and UAV43. Experiments show that the IOU loss method can reduce boundary loss and make STN module output spatial transformed image appropriately. Our proposed ST-BCNN model outperforms other advanced CNN models on all three datasets.

In computer animation applications, simplifying the mesh is an important technique to reduce the workload of visualization processing. In this paper, we propose a new method to simplify the dynamic model using local area and deformation information control. Our method uses an improved quadratic error metric with the local distortion measurement as the basic level. In addition, the deformation parameters are defined to be embedded in the summarized quadratic error, so that the regions with large deformations during the animation can be successfully retained. The new method is efficient and easy to implement, and in average can generate high-quality animated models with well-preserved local and deformation features on eacn frame.

The pantograph is an important component of railway pantograph-catenary system, which can provide electric current for electrified railway electric locomotive. Since the pantograph is in an open roof environment, the defects of the pantograph are inevitable in the long run. In order to ensure the safe operation of trains, in this paper, we propose a new method for anomaly detection of pantograph based on salient segmentation and generative adversarial networks. First, an object location model is trained by U-Net which perform excellent properties for a small number of samples to accurately extract salient area of the pantograph. Second, a generative adversarial model is constructed to generate the reconstructed salient images of pantograph by vector mapping. Finally, the structural similarity algorithm is used to evaluate the similarity between the input salient image and the reconstructed salient image, so as to extract the image difference and realize the anomaly detection of the pantograph. Experimental results validate the effectiveness and accuracy of our approach.

In this paper, we find that the overlapped detection frames of dense disease in the bridge disease data set, combined with the current research on target detection algorithms in dense scene. Based on YOLOv3, the feature pyramid is optimized to manually set thresholds for post-processing after extracting features. A bridge disease target detection model based on SE-YOLOv3 was proposed. In the post-processing process, IoU uses Soft-IoU to detect the accuracy of the prediction box, approves the value of IoU, and then uses the EM algorithm based on the Gaussian distribution to perform a function operation to delete the low-score prediction detection box to achieve the purpose of removing overlap. Experimental results show that the detection accuracy is 95.73%.

Aiming at the problem that the existing computer generated images (CG) and real photograph (PG) have low recognition rates, this paper proposes an improved convolutional neural network method to achieve accurate recognition of the two photos. Firstly, a convolutional neural network classification model is established, then different models are established based on the VGG-19 network structure, and transfer learning is introduced to reduce training time and save computing resources. Finally, a softmax classifier is used for classification. The experimental results show that the accuracy of the proposed method for the recognition of PG images is about 93%. Compared with other methods, the method has the highest recognition accuracy, which indicates that the method is feasible and effective.

Aiming at the unconscious lane departure behavior in the process of driving, a design scheme of lane departure warning system based on Android is proposed. STM32F103C8T6 is selected as the main control chip and CAN bus communication technology is used to collect vehicle information such as vehicle speed, steering angle and steering signal. Then Bluetooth module is used to transmit the data to the mobile phone for real-time processing. Taking GB/T 26773-2011 (Performance Requirements and Test Methods for Lane Departure Alarm System) as the design standard, the lane departure rate ε and the time T when the right front wheel crosses the lane line are considered as the input of the two-dimensional fuzzy controller, and the early warning level W is considered as the output. When the driving deviation angle is 5°, the interval of the lane departure rate ε is [0.75,1.35], and the interval of T is [0,1.5]. After normalized parameter processing, a certain fuzzy interval is set according to the driving habits of different vehicle drivers, and the fuzzy control algorithm with improved membership function is adopted for control, so as to improve the recognition rate of lane lines, reduce the early warning misjudgement rate and reduce the incidence of traffic accidents.

As the starting point of cutting, sawing is an important link in the production process. Circular saws have high precision and high sawing efficiency. Once a failure occurs, it will bring considerable losses. This paper presents a fault prediction method for circular saw based on state evaluation and ARIMA model. According to the mechanism of failure analysis, considering different parameter indicators for different failures, we first find out the components of the equipment involved in these parameters, and adopt the entropy weight method and trapezoid-triangular membership function model to determine the health status. The key mechanism of the components in poor condition is predicted using the ARIMA model, and the weighted method is used to predict the failure time.

With the development of China's industry and the "Made in China" and "Created in China" proposals, the automotive industry has developed rapidly. Nowadays, China's car ownership and sales volume has ranked first in the world. The rapid growth of vehicle ownership has also brought problems such as traffic congestion and frequent accidents. Traffic accidents often bring double losses of human and financial resources. Rear-end collisions should be given a lot of attention as high-incidence traffic accidents. This paper is based on the research of auto-tuning fuzzy control of vehicle auto-tuning system. Based on the vehicle's sensors to obtain the optimal road surface slip rate and road peak adhesion coefficient, the driver's input signal is synthesized, and the parameter auto-tuning fuzzy control is applied to multiple. After the fuzzy input is inferred from the complex input, the fuzzy output is determined to obtain the determined output signal effect and speed limit on the motor. At the same time, the safety distance of the actual road is predicted based on the peak adhesion coefficient of the road surface to remind the driver to drive safely and maintain the distance. This system can not only remind the driver to drive safely with a safe distance, but also limit the speed of the vehicle itself to ensure the safe driving of small cars.

In recent years, with the continuous development of the Internet and artificial intelligence, face recognition technology has also been widely used in many application scenarios. Facing complex surveillance scenarios, face recognition technology still faces great challenges. This paper focuses on implementing real-time and efficient face recognition systems in complex surveillance scenarios, such as insufficient lighting, small faces, dense crowds, and sides at 45 ° environment. The system is mainly based on RetinaFace for face detection and face alignment, and uses lightweight mobilenet (0.25) as the backbone network of RetinaFace. Facial feature extraction is based on deep residual neural network combined with ArcFace loss, and feature matching is performed by Euclidean distance. The experimental results show that the face recognition system has good real-time performance, accuracy and robustness.

In multiple object tracking(MOT), data association is a crucial part. By constructing similarity loss matrix for trajectories and detections, they can be matched correspondingly by using Hungarian matching algorithm. However, the similarity loss is often obtained by calculating the Euclidean distance or other handcraft distance metrics of features extracted between objects, which may not be robust enough, resulting in matching inaccuracy. In this paper, we propose a novel MOT method with applying deep learning to feature extraction and data association. We firstly design the appearance feature extraction network(AFN) to learn effective features by training it on a large-scale person re-identification dataset(reid). Then, we propose the similarity matrix estimation network (SMN) to obtain reliable similarity by training it on the public MOT Challenge dataset, MOT17. Additionally, the similarity matrix output of SMN includes the dummy objects, which are used to deal with the association problems of object missing and object appearance between frames. In the end, our proposed MOT method is evaluated on MOT15, MOT17 and ablation study is carried out.

In this paper, we propose an alongshore ship detection method based on multiscale feature fusion of rotation region proposal networks, the method provides an end-to-end ship detection framework, which is divided into three modules: multiscale feature module, rotation region proposal network module and context rotation region pooling module. The multiscale feature module integrates multiscale features, the rotation region proposal module can generate rotating bounding boxes in any direction, the context rotating region pooling module fuses the context information to pool the rotating region and complete the classification and position regression of bounding boxes. Experimental results show that our method can achieve good performance in remote sensing data sets.

This paper introduces a novel method to recognize the facial micro-expressions by directly adopting state-of-the-arts deep learning algorithms. Specifically, genetic algorithm (GA) is applied based on the principles of evolution in searching an optimal solution to generate discriminant features. Prior to that, the feature vectors of each video are first encoded using existing apex-based feature descriptors, viz, Bi-WOOF, OFF-ApexNet and STSTNet. Then, GA is utilized to enrich the features by eliminating irrelevant information that do not contribute to the expression prediction. It is acknowledged that the overfitting phenomena can be avoided in the feature selection process, as GA employs tournament selection and deterministic mutation procedures during the evolution procedure. As a result, Genetic Algorithm in Apex Frame Network (GA-ApexNet) is introduced and significant improvement of the recognition performance has been achieved. A standard experimental configuration is designed to evaluate the robustness of the proposed framework. Following the suggestion in Facial Micro-Expressions Grand Challenge (MEGC 2019), the UF1 and UAR obtained are 78.85% and 78.28%, for the composite dataset that comprises CASME II, SMIC and SAMM. We also note that this is the first work that fuses GA in micro-expression recognition system.

Objective: It is hoped that the DNN-Hidden Markov Model (HMM) acoustic model and the long-short-term memory neural network-Hidden Markov (LSTM-HMM) acoustic model can be used to improve the accuracy rate of speech recognition in keyword detection. Method: First, the principles of speech recognition and related algorithms are introduced. Then, a DNN algorithm is applied to the speech recognition system, and a keyword detection system is established based on the DNN-HMM acoustic model. As for the experimental comparison, the proposed DNN model is used for acoustic modeling, and the effect of the DNN algorithm on the performance of the recognition system is analyzed experimentally. Results: The size of the training parameters of the proposed LSTM model is 436570, the size of the training parameters of the deep neural network (DNN) is 698100, and the size of the training parameters of the Gaussian mixture model (GMM) is 1226700. The accuracy of speech recognition based on the LSTM-HMM model and DNN-HMM model is 965% and 91.6%, respectively, which is significantly higher than 78.5% of the traditional speech recognition model (GMM-HMM). Conclusion: The speech recognition technology based on LSTM-HMM and DNN-HMM models has higher accuracy and is more suitable for speech keyword detection.

UAV flight control systems have the characteristics of non-linear, multi-variable, and strong coupling. Traditional PID control algorithms have many problems in complex flight environments, such as large adjustable parameters, poor robustness, and slow convergencein. Due to the defects of traditional PID control algorithm, this paper proposes an improved PID control algorithm based on fuzzy neurons. The modified fuzzy neuron is used to modify the traditional PID control algorithm. Through Matlab experimental simulation, it is proved that the algorithm has a significant improvement in response speed, robustness, accuracy and anti-interference compared with the traditional PID algorithm.

Because the temperature control system of glass furnace has the characteristics of non-linear, time-varying, large amount of interference, large lag and so on, the traditional PID control is difficult to ensure the temperature control accuracy and can not achieve the expected effect, the artificial intelligent control is proposed for the temperature control system of glass furnace. Humanoid intelligent control is to make the controller imitate the control behavior of human. It uses control theory and computer technology to act on the actual control system. Through the analysis of the simulation curve of the artificial intelligent control of glass furnace temperature, the superiority of the artificial intelligent control is verified.

Through the analysis and in-depth study of the fuzzy control system model and neural network, the optimal control strategy is obtained by the organic combination of fuzzy reasoning and neural network analysis training, and the automatic control algorithm based on fuzzy neural network is designed. The modified algorithm is applied and verified in Modern Agriculture Demonstration Park of Chengdu Agricultural College. The results show that the designed algorithm can meet the demand of automatic control in the field of intelligent agricultural control, and the control system has good universality and scalability. The application of automatic control algorithm based on fuzzy neural network can accelerate the development of agricultural modernization in China, and lay a solid foundation for the further research on the application of automatic control in other fields.

A modified Hopfield neural network algorithm is proposed and applied to the path planning to solve some problems in the traditional Hopfield neural network. Firstly, the traditional A* algorithm is used to select the nodes in the search area that meet the criteria. Then, the nodes conforming to the standard are converted into neurons in the Hopfield neural network, and the stability of the network is used to iteratively select an optimal path. Experiments show that the improved Hopfield neural network algorithm can reduce the search time of path planning and improve efficiency.

MIMO (multiple input multiple output) vehicle anti-collision radar refers to the application of MIMO system radar in vehicle anti-collision. MIMO radar is a new type of radar which has been put forward in recent years and attracted great attention. It is a kind of radar with multi transmitting and multi receiving. The characteristics of wave form diversity make it have many performance advantages, such as strong anti-jamming ability and strong resolution, compared with traditional radar. In recent years, many researchers have studied the application of MIMO Radar in the field of aviation and land transportation and made remarkable achievements. Compared with the traditional radar, the advantage of MIMO radar is due to the orthogonal performance of the transmitting waveform. The related problems of the pattern are deeply related to the resolution. As an important reference index in radar, the resolution is worthy of further study. In view of the above problems, the main work of this paper is as follows.

With the fast development of automatic driving and video monitoring applications, video dehazing is a vital problem in computer vision because turbid images can have a large impact on their performance. As one of the remarkable techniques, convolutional neural networks have been greatly developed and shown great effectiveness in video dehazing at present. Inspired by deep learning, this paper introduces one of the convolutional neural networks named AOD-Net and collects a dataset to evaluate its dehazing performance. By comparison, the proposed algorithm achieves high SSIM and high PSNR on both our collected dataset and NYU Depth V2 dataset. AOD-Net not only generates well-dehazed images, but also learns very quickly during the training process. This All-in-One Dehazing Network improves visual quality of hazy images to a large extent.

Defect detection methods based on machine learning extremely accelerate the substation routine inspection process. In this paper, we propose an automatic defect detection method based on improved Faster RCNN. For one thing, random feature pyramid (RFP) structure is introduced for the highly discriminative feature map construction; for another thing, we execute the detection boxes selection by soft non-maximum suppression (SNMS), keeping the detection of defects which distribute densely. Finally, online hard example mining (OHEM) is employed to deal with the imbalance problem. Experimental results demonstrate that the proposed approach obtains competitive performance compared with state-of-the-art deep learning object detection methods.

with the Explosive Development of Deep Learning Technology Face Recognition and Other Recognition Technologies Mostly Adopt Deep Learning Algorithm for Recognition. Although the Deep Learning Algorithm Has High Recognition Accuracy, It Has a Huge Demand for Computing. in the Mobile Terminal, We Can Use Artificial Intelligence Chips That Can Accelerate Deep Learning Operations to Complete Relevant Operations. Deep Learning Has Fixed Modes, Like Convolution. Ai Chips Can Significantly Improve the Efficiency of Deep Learning Operations by Optimizing the Corresponding Operation Modes. in This Way, Mobile Terminals Can Quickly Implement Complex Deep Learning Operations, Such as Face Recognition Based on Deep Learning. One Representative of Ai Chips is the Tensor Processing Unit of Google, Which is Able to Accelerate the Tensor Flow of the Deep Learning System, Which is Far More Efficient Than Gnus. the Tpu Provides 1,530 Times the Performance Improvement and 3,080 Times the Efficiency (Performance/Watt) Improvement over the Same Cpu and Cpu. Traditional Face Recognition Algorithms Include Face Recognition Technology Based on Pca(Principal Components Analysis) and Face Location Technology Based on Ad Boost. Although the Traditional Face Recognition Technology is Fast, the Detection Effect is Much Different from the Deep Learning Technology. on the One Hand, the Accuracy of the Traditional Face Recognition Methods Represented by Pca is Far Lower Than That of the Deep Learning Algorithm. on the Other Hand, for the Recognition of Massive Users, the Traditional Pca Face Recognition Technology is Not Competent.

Image Recognition is the Use of Computers to Process, Analyze and Understand Images to Identify Different Patterns of Objects and Images. in This Paper, We Introduce the Idea of Nonlocal Processing on the Basis of Regularization Tv. the Process of Image Recognition System is Divided into Five Parts: Image Input, Preprocessing, Feature Extraction, Classification and Matching, in Which Preprocessing Can Be Divided into Image Segmentation, Image Enhancement, Two Value Reduction and Refinement and So on. the Paper Presents Research and Application of Multiscale Representation and Regularization Method in Image Recognition. At the Same Time, through Programming in the Matlab Programming Environment to Achieve the Relevant Algorithm, through the Actual Image Recognition Experiments, the Use of These Conclusions, and Finally to Accurately Identify the Results.

with the Development of Urbanization, It is More and More Urgent to Improve the Reliability and Stability of Power Supply. Due to the Consideration of Municipal Planning, Environment, Personal Safety and Other Factors, the Utilization Rate of Power Cables in Urban High-Voltage Transmission Channels Has Increased Dramatically. Accidents Such as Fire, Large-Area Power Failure and So on, Which Are Caused by Internal Insulation Defects of Cables, Defects in Joint Position Fabrication and Installation Process, Occur from Time to Time. This Paper Studies the Detection Method of Potential Insulation Defects in the Cable Interior and Joint Position Based on the Ac Withstand Voltage Partial Discharge Detection Technology, and Designs a Set of Distributed Partial Discharge Test System Based on the High Frequency Pulse Method to Detect the Interior, Middle Indirect Head and Terminal Head of the Cable. in the Field, the Distributed Pd Detection Technology Based on Ac Withstand Voltage Designed in This Paper Can Effectively Detect the Internal and Joint Insulation Defects of Power Cables, and Realize Fault Location, Which Has the Value of Practical Application and Promotion.

Based on the Encoder-Decoder Framework, Combined with the Pooling Preprocessing and the Mupli-Attention Mechanism, This Paper Proposed the Mp-Matt Model for the Prediction of Mine Gas Time Series Data. the Model Had a Good Predictive Accuracy in the Data Set of Oxygen, Carbon Monoxide and Methane in a Mine. the Accuracy of the Best Baseline Comparison Model Had Increased by 10.30%, 22.26% and 20.02%, Respectively. This Study Compared the Effects of Each Component in the Model. the Comparison Showed That the Multi-Attention Layer Had the Most Obvious Effect. the Increase of the Metic Pooling Layer Was Lower, and the Effect of Adding the Fatt Layer Was Slightly Better Than That of Adding the Hatt Layer.

Pedestrian attributes recognizing (PAR) is an important task in computer vision area due to it plays an important role in video surveillance. On the other hand, pedestrian visual attributes are treated as middle-level semantic features which can provide calibration information for high- level human related visual tasks in order to improve the discriminative ability of the models, such as pedestrian detection, people tracking, person re-identification, action recognition and scene understanding. However, the most study of PAR is based on single person. In this work, I implements a multiple pedestrian attributes recognition model based on the offline retail scenes. This model combines object detection and multitask classification techniques and can be trained end-to-end directly for back propagation. This paper also demonstrates the performance of the final model through a series of experiment.

With the continuous development of artificial intelligence technology, people's research on it has become more in-depth and machine learning has become a new research core issue in artificial intelligence. Before artificial intelligence technology was mature, it was a very difficult task to realize automatic character recognition and input. At present, human beings have made some progress in intelligent robots, voice recognition and network search. The text recognition method based on machine learning is of great significance to the promotion of information technology. Text recognition based on machine learning, although it has formed relatively rich theoretical support in principle and skill, but still faces various factors during actual practice, making its recognition effect difficult to be reflected. Based on the feature fusion perspective, this paper combines the background and main identification methods of text recognition in artificial intelligence machine learning to analyze the purpose, which lays a foundation for future specific applications.

With the development of social economy, intelligent transportation system (ITS) has been booming. It is mainly to realize the omni-directional, real-time, accurate and efficient guidance and control of transportation in a large range. Traffic flow prediction, especially short-term traffic flow prediction, is the basis of urban traffic control and guidance, so traffic flow prediction system is one of the important subsystems in its. Short-term traffic flow prediction technology belongs to the important research field of intelligent traffic control and vehicle guidance. It can help cities to conduct intelligent traffic guidance and enable users to choose the optimal path. Prediction of road traffic flow state and auxiliary guidance are very important basic theories. Based on the theory of artificial intelligence, this paper analyzes the short-term traffic flow prediction and auxiliary induction methods, points out the limitations of the existing prediction methods, and proposes a short-term traffic flow intelligent prediction method combined with artificial intelligence technology.

Based on the new environment of big data, this paper expounds the connotation and characteristics of big data, and analyzes the characteristics of consumer behavior under the application background of big data analysis technology. With the help of AISAS model, which is used to analyze consumer behavior in the network economy, and in combination with the influence mechanism of big data analysis on consumer behavior decision-making process, we have constructed a consumer behavior model under the background of big data and tested it by means of questionnaire survey. The results show that the factors that affect consumer decision-making include external factors and internal perception, while big data affects consumer internal perception through the impact of external factors, thus affecting consumer decision-making. At the same time, consumer information sharing is also conducive to improving the accuracy of big data analysis.

With the increasing demand for high-precision positioning of civilian positioning terminals, the combination of single-frequency global navigation satellite system(GNSS) Real-Time Kinematics (RTK) and Inertial Measurement Unit of the Micro Electronic Mechanical System(MEMS-IMU) has been selected for high-precision positioning in the industry with its advantages of low cost, miniaturization and high precision. Based on the extended Kalman filter single-frequency RTK and Ultra low-cost MEMS-IMU loose combination positioning algorithm, the data is collected and tested in static scenes, and dynamic scenes. The experimental results show that: (1) In the static scenes, the combined positioning accuracy is better than 1.5cm, which is equivalent to the accuracy of the dual-frequency commercial receiver; (2) In the dynamic scenes, the single-frequency GNSS RTK positioning accuracy is better than 3cm, and the combined positioning accuracy is better than 5cm. In general, the combined positioning in static and dynamic modes can provide reliable high-frequency centimeter-level positioning.

Understanding the action text is the main problem of dynamic scene generation in text-to-scene conversion. However, the existing basic research about it is still lacking, and its semantic parsing method is not perfect. Therefore, based on the idea of action chain structure, the formal expression of it is established on the basis of defining action text, classifying action text and analysing the features of action text. The method of action text semantic parsing is proposed combing with the action text semantic dictionary. The experimental results show that the method completes the semantic parsing for the action text to a certain extent and provides a reference for computer to understand the semantic of action text.

Under the support of modern technology, Internet finance as a new financial model gradually improves the work structure and changes the traditional forms of financial work, so that the work becomes more convenient. After the emergence of Internet finance, due to the short time of its development, we should also introduce modern technology in the process of carrying out related work according to financial management, and use big data to analyze working data, and find that many work will be difficult to carry out because of lacking of scientific management mechanisms during the period of Internet financial application so that it cannot achieve the desired effect. There are great security risks in the Internet financial market because of lacking of credit risk management. In order to improve our financial system, we need to pay more attention to the credit risk management. According to the problems existing in the operation of Internet finance, we put forward some solutions.

This paper describes the research background of malicious domain name detection based on DNS data analysis, obtains DNS data through active domain name data or passive domain name data, then obtains DNS data through knowledge-based methods, detection methods based on machine learning and hybrid methods, and finally puts forward research suggestions.

As many finance services are made through supply chain system, it needs the stability and accuracy in information transmit process. Supply chain finance service online should be built on the system, including open API, basic database, and progress standardization. The paper focuses on related processes on the SCM platform. Then it discusses how to strengthen the determination of legal rights and responsibilities and provides business scenario of planning electronic warehouse receipt. It shows that the information transmit mechanism can get the economic profit with the help of system.

The rapid construction of satellite systems has become a priority development direction of space technology. Space plug-and-play avionics (SPA) technology has the characteristics of short development cycle, low cost, strong pertinence, and quick results, so it has become a hot research topic in various countries today. This paper comprehensively summarizes the technical system and development trend of SPA. Firstly, the SPA hardware architecture and its three generations evolution process is analyzed; then, the application of SPA technology, including the development of PnP series satellites, OSR-1 satellite, panel extension satellite is introduced; finally, the development trend of SPA technology, including coordinated development of hardware and software technologies, and transforming traditional satellite test methods is also put forward.

According to the characteristic that structural health monitoring have many monitoring sensors and monitoring signal types, with DSP and FPGA used as the core controller, a data collection system was designed. The designed system can realize multi-channel parallel data collection, flexible port configuration and good scalability, which meets the requirements of high-speed collection and processing and real-time online monitoring. The designed data acquisition system has the advantages of high performance, low cost and convenient application. In addition, the system has a wide sampling frequency range, which can adapt to a variety of sensor signal sampling requirements, is easy to install, easy to achieve network configuration, and can better meet the structural health monitoring of railway vehicles.

There are many redundant attributes for the original data of the emergency system with uncertainty and data loss, and the existing incomplete information system attribute reduction algorithms have the disadvantages of high time complexity and completeness of the reduction result, an attribute reduction method for incomplete emergencies based on finding the maximum mutual information is proposed. Firstly, the original data of the emergency was preprocessed, and the situation and the consequences were regarded respectively as conditional attributes and decision attributes, thereby an incomplete decision information table was constructed; then, according to the characteristic that the mutual information of the conditional attribute set and the decision attribute is equal to the information entropy of the decision attribute in the consistency incomplete information system, the attribute reduction algorithm based on finding the maximum mutual information was obtained; finally, this method was used to reduce the attribute of coal mine fire data and compared it with other three methods. The results indicate that this method can effectively reduce the redundant attributes of coal mine fire emergencies, and can ensure that the reduction results have strong completeness when the algorithm time is low, thus decision-making for the emergency department is provided.

Through the use of agricultural intelligent sensing equipment and intelligent control equipment, the precision planting management system builds the Internet of Things monitoring network in the agricultural production site to realize real-time data collection, monitoring of the climate environment, soil conditions, crop growth, pests and diseases, and remote automation control of various agricultural facilities and equipment on the site. The collected agricultural big data is uploaded to the planting management cloud platform in real time, and statistical analysis is performed instantly, which addresses the staff's demand for data collection and analysis. Through data mining and analysis, the rules and expertise of high-quality and high-yield crops are obtained, which further guides the regulation of agricultural varieties and production environment, so as to achieve intelligent planting.

With the increasing amount of data, the in-depth application guided by data prediction has ushered in the climax of development However, due to the limited depth and breadth of the data owned by each unit, the data among different units are not common, so more and more data islands are formed in the city. Data sharing has become a major obstacle to the construction of smart city. To solve the problem of data sharing, in 2016, Google first proposed the concept of Federated learning. However, the bandwidth of wireless communication has not increased significantly. As a result, the bottleneck has shifted from previous computing to today's communication problems. During my internship in a company, I participated in a project to provide platform support for federal learning. The purpose of the platform is to facilitate the transformation of federal learning academic achievements into commercial products. Based on this project and the current situation of federal learning, this paper summarizes the secure communication problems of federal learning.

In view of the deficiency of domestic Weibo user attribute analysis, the imperfection of Weibo feature extraction and the problem that the classification accuracy needs to be improved, a Weibo user attribute analysis method based on multi-features is proposed. This paper first uses the word2vec model to build text features from Weibo content, then constructs Weibo user features from Weibo information and user information, and finally sends multi-feature sets into the improved three-tier stacking model to build Weibo user attribute analysis model. The experimental results show that this method has better classification effect than the text-based classification method and the traditional stacking model.

In recent years, most of Lora's research in the field of wireless technology has focused on the experimental exploration of its stability and communication range, without much attention to the impact of the actual transmission rate. However, as the length of Lora transmission load increases, the growth rate of low-speed transmission time is faster than that of high-speed transmission time. Therefore, it means that for the transmission of large capacity packets, the lower the transmission rate, the more time it takes; the longer the time, the more power it consumes. In this paper, the SNR, RSSI and packet loss rate of Lora over 3 Kbps in urban area and mountainous area were tested and explored respectively, and the RSSI curve with distance in different environment was obtained. The test results show that the actual communication distance of Lora with high transmission rate of 21875 bps in line of sight environment can reach 3 Km, packet loss rate is less than 1%. In addition, the influence of Lora on communication quality under different occlusion degrees in a certain range was tested. In the range of 500 m to 3 km, when the occlusion ratio was 0-5%, the highest speed reference value was 21875 bps, 5-10% reference value was 6250 bps, and more than 10% reference value was 3125 bps. In this paper, Lora SX1278 transceiver was used, in the same transmission of 100 KB data, the actual transmission rate is further improved by using the maximum duty cycle and improving the request response mechanism. Compared with a Lora system for image transmission, the transmission time is reduced by 81.3% and the transmission speed is increased by 4.3 times, which greatly improves the transmission efficiency of large capacity data such as image.

With the development of aerospace technology, space security has become the focus of national attention, and space offensive and defensive technologies have become increasingly important. This article summarizes the attack methods of space satellites and aircraft. It focuses on analyzing three types of space attack technologies: kinetic energy attacks, directed energy attacks, and communication interference. In-depth analysis of anti-satellite missiles, common-orbit satellite interceptors, high-energy lasers, microwaves, particle beams, communication suppression, communication deception and other attack technologies, introduced the maturity and test situation of each technology, and sorted out for further development of space attack and defense technology research Several research directions.

The lip region provides the most direct visual information in the process of multi-sensory speech perception, which is applied to speech recognition and lip reading. In this paper, we extract eight lip features in articulating the basic vowels [a], [e], [i], [u], [ü] in standard Chinese, and analyze the efficiency in distinguishing the five vowels combined with articulatory phonetics. We use Dense Convolutional Network (DenseNet) to process two-dimensional lip images and fuse the lip features to identify the Chinese with consonants. The results show that the application of lip shape features in Chinese vowel recognition and Chinese consonant lip reading is consistent. Two-dimensional lip images can effectively improve the recognition rate by fusing lip features in lip reading.

With the development of space-based technology, the construction of the space-based network has been into a period of major change, countries are putting forward their own space-based network and getting them off the ground. This paper analyzes the space-based network development situation, introducing the SCaN network in the United States, Europe ISICOM plan and the development of China's SGIN, discussing the trend of the development of the space-based information network. At present, the construction of space-based network is still based on the foundation network as the core, the space-based system as the expansion, based on the communication interruption satellite, gradually evolves into the space-based network, in the future it will integrate the space-based navigation, space-based information acquisition to form a space-based integrated information network.

In the current network, there are still many security issues, such as identity fraud and address spoofing. In traditional Public Key Infrastructure(PKI), most trust models adopt a centralized trust model with a single-point trust anchor. The existing PKI system lacks a unified authentication framework and is low efficiency. Aiming at the problems of how to store public keys efficiently and how to make the network secure and trustworthy, this paper proposes a new trust anchor framework based on consortium blockchain, named Trust Anchor Chain(TAC). This study uses the form of consortium to manage TAC, proposes a hierarchical naming ID, named TAC_ID to uniquely identify the user in TAC, and proposes a new data structure, named Struct Radix Trie, to store public key information in TAC to improve efficiency. Section 4, TAC has superiority through the comparison between local experiments and remote experiments. In TAC, users can access each other in the TAC, get the public key, and ensure the correctness of the identity. All users work together to maintain TAC to achieve equal network interconnection, security and credibility.

The operational test, along with the evaluation based on it, is the most important decision-making basis for the Department of Defence in acquiring weapons and equipment. However, there has been relatively little attention to the credibility of its conclusions. In order to evaluate the decision-making risks quantitatively, the relationship between the two types of errors in the operational tests is studied. By analysing the interaction between the factors affecting decision risk errors, it is pointed out that it is of great practical significance to ensure that the tests have sufficient statistical power, which is usually overlooked before in operational tests.

This study selected the five typical Wells area as the research object in the town of Li Qu, Baota district, Shaanxi province. The petroleum hydrocarbon content of soil samples collected from 0~10m, 10~30m and 30~50m outside the oil well center was studied and analyzed, the results showed that the well surrounding soil samples of petroleum hydrocarbon concentration in the range of 73-1090 mg/kg, much higher than that of petroleum contaminated soil (500 mg/kg), and petroleum hydrocarbon concentration are radiation pollution radius distribution. According to the standards of oil contaminated soil remediation stipulated by Oklahoma (500mg/kg) and Canada (450mg/kg), the soil around the oil contaminated soil remediation needs to be remediation in different years of production. In addition, due to the specific degradation effect of indigenous microorganisms on petroleum hydrocarbons in the soil, the soil pollution level of the oil Wells in operation is higher than that of the abandoned ones. The soil oil pollution in 2 # and 5 # oil wells is about the critical value in the soil remediation standard.

Based on the research of lean production, the lean supply theory is a development of the lean concept from the production field to the supply field. After that, it is developed integrated lean supply which is a kind of collaboration in the first level and provides components with higher quality and lower cost for assembly plants in larger size, higher level, shorter time and enterprises in the supply chain for a long term and it could finish the process of supply with higher lean requirement on the premise of highly integration. The supply chain system in auto-mobile industry is very complicated. At the same time the operational efficiency of supply logistics will directly influence the final profit of enterprises in the supply chain. However, in the process of implementing lean supply logistics, due to the lack of specific and systematic theoretical guidance, the research results are not satisfactory. This article base on the perspective of collaborative innovation, and through collaborative innovation of supply-chain, production and distribution logistics in different stages of logistics operation to realize logistics integration of a enterprise successfully.

Aiming at the problem that emergency escape parachute cannot meet the requirements of the index using the binomial evaluation method, combined with the characteristics of emergency escape parachute products, comprehensively using emergency escape parachute components and system test data, the emergency escape parachute Bayesian reliability assessment method is proposed to solve the problem of reliability assessment in the case of small sample size of emergency escape parachute system test, which effectively reducing emergency escape parachute test expenses and development cycle.

This paper proposes a vehicle object assoication method based on millimeter wave radar detection, which can improve the accuracy of driving assistance functions that take vehicle targets as input. This method is used to judge the data association of the target tracking group through Kalman filtering. The L-shaped model is used to initially determine the outline of the measurement object, and the close object clusters are segmented according to the standard vehicle parameters and the related speed. Furthermore, a Gate algorithm is used to associate the hypothetical object within a certain area with the target. Road test experiments show that the data assoiation method has high accuracy for detected vehicles.

The importance measurement of nodes in directed weighted networks is of great significance for theoretical research and practical application of networks. Aiming at the directed weighted network model in actual networks, based on the special transformation of graphs and tree roots, this paper proposes a comprehensive quantitative assessment method for the importance index of nodes based on the directed weighted network's global properties, local attributes, network location and bidirectionally propagation. This method transforms the nodes to be analyzed in the graph to root, constructs a bidirectional depth-affected XD tree and a breadth-affected XB tree with the node to be evaluated as the root of the tree. Then comprehensively consider the two-way influence of the node level, node neighbors and the closeness between neighbors in the two trees constructed, and the in-out effects, build a deep impact evaluation model at the two-way level, two-way neighbor information and comprehensive evaluation index of clustering coefficient. By calculating the comprehensive importance value of each node in the network, the comprehensive ranking of the importance of the nodes in the network can finally be obtained. The robustness experiment of ARPA network shows that the G value of the important node quantified by this method has absolute advantages, this method can effectively analyze key nodes in the network.

The traditional end-to-end task-oriented dialogue models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. But in the case of large amounts of data, there are many types of questions. It performs poorly when answering multiple types of questions, memory information cannot effectively record all the sentence information of the context. In view of the above this, this article uses a modified transformer model to overcome the problems mentioned in dialogue tasks. Transformer is a model constructed using attention mechanisms, which completely discards the method of RNN (recurrent neural networks), and its structure includes two sub-parts of Encoder and decoder. It uses residual network, batch normalization, and self-attention mechanism to build the model structure, uses Positional Encoding to capture sentence information, which can speed up model training convergence and capture Longer sentence information. In this paper, we modified the activation function in the transformer and use label smoothing to optimize the training to make the model's expressive ability better than previous.

Aim: Exploring the impact of nephropathy patient characteristics on length of stay (LOS) grading, proposing a path of LOS classification based on the characteristics of patients, and providing suggestions for the accurate management of shunting of patients with nephropathy patient to promote the sustainable development of the hospital.

Methods: The data of inpatients from the Department of Nephrology of a large hospital in 2016 were used, including five variables: gender, age, patient type, medical insurance type, and LOS. Based on quantifying patient attribute variables, We use three steps to finish the grading. Firstly, using the factor analysis to extracte the common factors of patient characteristics. Sencendly, according to the results of factor analysis, using k-means clustering analysis to classify the patients. Finally, According to the characteristics of different types of patients and the law of LOS differences, a LOS classification path based on patient characteristics is proposed.

Results: The factor analysis shows that the LOS common factor characteristics are disease characteristics, attribute characteristics and reimbursement ratio characteristics. The k-means clustering indicates that the patients are divided into 5 categories: the mean LOS in category 1 is 15.78, Patient characteristics: Mostly elderly women with the blood resuscitates patients(38.2%) or tumor recovery patients(30.3%), city medical insurance(50%);the mean LOS in category 2 is 10.5, Patient characteristics: Mostly strong men with the ordinary patients(62.5%), City medical insurance(79.2%);the mean LOS in category 3 is 7.62, Patient characteristics: Mostly young men with the other patients(99.7%), Provincial medical insurance(73.1%);the mean LOS in category 4 is 13.7, Patient characteristics: Mostly women in pre-old age with the Ward daytime patient(38.9%) or other patients(31.8%), Urban rural medical insurance(60.6%);the mean LOS in category 5 is 6.73, Patient characteristics: Mostly young men with the other patients(99.3%), Provincial medical insurance(54.4%).According to the characteristic differences among patients and the law of LOS differences, a model of patients' LOS classification path was proposed.

Conclusion: The LOS classification path based on patient characteristics can realize the pre-classification management of patients, which has practical significance for early intervention of hospital resources.

As an important branch of intelligent robots, brain-computer interface (BCI) technology has great application prospects in the fields of neurological rehabilitation, functional prosthesis assistance and neurological monitoring. Many breakthroughs have been made in the research of brain-computer interface. Based on the existing research literature, this paper has done the following work: outlined the research significance and development status of brain-computer interface technology; introduced the working principle and structure of brain-machine interface system; compared EEG signals of different frequencies, and classified their application fields; analyzed five preprocessing methods, and listed their advantages and disadvantages; summarized the theories and algorithms involved in feature extraction and pattern classifier. Finally, after a detailed introduction of the non-intrusive brain-computer interface components, the paper summarized the current technical challenges faced by the brain-computer interface system.

Electroglottograph(EGG) is a physiological signal collected from the throat which reflects the vocal cord movement. EGG signals can be still collected from patients without speaking ability or from the extremely noisy environments. Additionally, the trends of the vocal cord movement will be distinctive for long enough Chinese sentences with different contents. Therefore, it is valuable and possible to carry out the research of applying only the EGG signals for content classification or recognition. In this paper, a content classification method with EGG was proposed, which consists of an EGG feature extraction module and a classification network based on LSTM(Long Short-Term Memory) units. The EGG feature extraction module was composed of three parts: the voiced segments extraction, the feature extraction and the F₀ smoothing. The classification network was made of a three-layer bidirectional LSTM encoder. This method achieved 91.12% accuracy on the validation set in 20-class content classification experiment, which provides the reference for further study in content classification and recognition with EGG signals.

PSOD process has serious hysteresis phenomenon and belongs to the large hysteresis system. On the basis of analyzing the working mechanism of the system, the PID control, Smith predictive control and fuzzy control are combined to control the system. Simulation results show that the proposed method has better dynamic performance than the traditional PID control method, regardless of the given input or interference. The conclusion is valuable for the safe operation and control of time-delay systems.

It Introduces the Concept of RFID Technology and Logistics Distribution. According to Their Characteristics, It Designed and Studied a Logistics Distribution System with Its Overall Scheme, Hardware and Software. the System Can Identify and Track the Information of the Whole Acquisition and Supply Chain. It Can Provide Accurate Product Distribution Information for Suppliers, Manufacturers, Retailers and Consumers. It Can Greatly Save Time and Reduce Labor Costs, Simplify the Process of Logistics Distribution and Improve the Efficiency of Distribution.

Asp is a Script Running Environment on the Server side. through This Environment, Users Can Create and Run Dynamic Interactive Web Server Applications, Such as Interactive Dynamic Web Pages, Including the Use of Html Forms to Collect and Process Information, Upload and Download. This Website is a Music Education as the Theme of the Website; This Paper Will Introduce the Website Material Organization, Production Process ". the Characteristics of Part of This Web Site Are the Online Message Function. It Includes: Message Board Application, Messages, Browsing, Moderator Management Functions. All Data Read and in through the Establishment of the Access Database Server. the Main Page Using C# Language to Operate the Database, to Achieve through the Database Query, Form and Data Verification Using Java Script Scripting Language, and to Control the Page Display Correctly the Type of User Authority Design. Using Sql Database Contains Several Tables, to Realize the Operation of the System.

In order to solve the problem of low performance caused by errors and omissions in tags in the data resources, this paper proposes a tag optimization method based on self-attention generative adversarial nets(SAGAN). First use TF-IDF training to calculate the similarity between texts, then use the consistency between the similarity between texts and tag similarity to define the objective function. Second add correction terms to reduce the bias of user-provided tags before and after optimization. Finally, we propose to apply Self-Attention GAN to further improve the performance of tag optimization objective function, and the results are compared with the original tags. Compared with the original tag, the optimized tagging performance has been improved. The test results show that the tag is optimized and improved, and the performance problem in tag recommendation is solved.

As for the classification network that is constantly emerging with each passing day, different classification network as the backbone of the semantic segmentation network may show different performance. This paper selected the road extraction data set of CVPR DeepGlobe, and compared the performance differences of VGG-16 as the backbone of Unet, ResNet34, ResNet101 and Xception as the backbone of AD-LinkNet. When VGG-16 is used as the backbone of the semantic segmentation network, it performs better in the face of long and wide road extraction. As the backbone of the semantic segmentation network, ResNet has a higher ability to extract small roads. When Xception is used as the backbone of the semantic segmentation network, it not only retains the characteristics of ResNet34, but also can effectively deal with the complex situation of extracting target covered by occlusions.

Taoist culture has profound guiding significance to people's spiritual life and social life, so it is realistic and necessary to inherit and develop Taoist culture. Language is an important medium for the spread of Taoism, and its two main forms have their own characteristics. The emergence and rapid development of network technology have promoted cultural exchanges and brought new opportunities for the inheritance and development of local culture. It is necessary for us to discuss the advantages and disadvantages of the application of network technology in Taoist language communication, so that computer network technology can be better applied in Taoist language communication practice. With the help of the advantages of network media, we can build scientific and effective cultural communication strategies, such as establishing communication awareness, expanding the scope of communication and increasing the forms of communication. The continuous development of information communication is the development of information technology. This paper analyzes the characteristics of Taoist language communication media under the background of the rapid development of computer network technology.

The research aims to analyze students' psychological mood during online study and enhance the efficiency of learning. In this research, face features are extracted by LBP (local binary pattern) algorithm based on face feature recognition. The facial feature points are tracked by CLM (constrained local model) algorithm, and the face images are normalized by AdaBoost algorithm. In the end, the fatigue level of learners is calculated by P80 method in PERCLOS criterion to verify the accuracy, recognition rate and error rate of the model. The results indicate that through the recognition of the faces in a complex background, the accuracy of the mode is 95%, the recognition rate is 95.53% and the error rate is only 0.53%, the range of aspect ratio in the human eye image is 0.22 ≤ λopen ≤ 0.27, 0.05 ≤ λclose ≤0.1, and learner's level of excitement can be estimated according to the range of f. Therefore, this algorithm model can extract learner's facial features very well and show a good result. Detecting the degree of learner's excitement accurately can provide an important fundamental for the realization of online learning emotion detection system which also has guiding significance for the development and popularization of the networked education.

The economic vitality of China's regions (provinces or cities) is affected by many factors. In this paper, the influencing indicators are listed by referring to the relevant factors that affect economic vitality. Based on the data obtained from the Internet, a systematic (hierarchical) clustering model of relevant indicators in Chinese provinces and cities is established, introducing the Elbow Rule and Factor Analysis to analyze and summarize the correlation of various indicators. A computer-based multiple regression model is used to establish a relationship model that affects economic vitality. Based on this relationship model, effective and reasonable suggestions for improving economic vitality are proposed.

The perfect computer simulation legal mechanism has an important safeguard function and plays a key role in building a socialist harmonious society. Chinese current legal system is still imperfect in terms of legislation and law enforcement, such as lagging legislation, unfair legislation, unfair justice, etc. Building a harmonious society is a topic of common concern to the whole society and is widely recognized by the public. Machine learning is an algorithm that computer programs can make judgments and inferences through learning data, thus simulating human intelligence activities and improving the performance of the system itself. The application of machine learning has brought opportunities for social development and injected vitality into legal reform. Therefore, based on computer simulation, this paper introduces the important role of perfecting computer simulation legal mechanism in building a socialist harmonious society, and elaborates four kinds of computer simulation legal mechanisms in building a socialist harmonious society.

How to identify and analyze influential nodes of Tactical Mobile Ad Hoc Networks is an urgent problem. Existing researches fail to make full use of features of Tactical Mobile Ad Hoc Networks. In this paper, we propose an influential nodes analysis method of Tactical Mobile Ad Hoc Networks based on impact factors evaluation (TINE). The method evaluates node importance through two mechanisms: connectivity analysis and node importance evaluation. In connectivity analysis, the current network topology is reconstructed by placing nodes to monitor network data packets and resolve protocol. In node importance evaluation, node importance degree is calculated by adjacency matrix. The experimental results show that TINE we proposed has advantages over the existing methods in Tactical Mobile Ad Hoc Networks.

Under the current economic situation in China and the world, it is impossible to promote sustained and effective economic growth by relying solely on the demand-driven model driven by exports, investment and consumption. Supply-side reform is a new normal for economic development. It is an important decision made by the Party Central Committee in the face of weak demand and a downward trend in the economy. It has defined the strategic focus of a new round of comprehensive reform. At present, the macro-economic problems facing our country have both cyclical characteristics, but they are mainly the trend problems brought about by structural contradictions. Promoting the structural reform on the supply side is a theoretical sublimation of our country's macro-economic environment and is also an adverse force of the current economic situation. Based on the computer simulation analysis of the connotation of supply-side reform, this paper sorts out the relevant researches on supply-side reform at home and abroad. Explains the necessity of supply-side reform in our country, and puts forward the path choice for our city to promote economic growth by supply-side reform.

This article has been retracted by IOP Publishing following an allegation that raises concerns this article may have been created, manipulated, and/or sold by a commercial entity. In addition, IOP Publishing has seen no evidence that reliable peer review was conducted on this article, despite the clear standards expected of and communicated to conference organisers.

The authors of the article have been given opportunity to present evidence that they were the original and genuine creators of the work, however at the time of publication of this notice, IOP Publishing has not received any response. IOP Publishing has analysed the article and agrees there are enough indicators to cause serious doubts over the legitimacy of the work and agree this article should be retracted. The authors are encouraged to contact IOP Publishing Limited if they have any comments on this retraction.

Retraction published: 23 September 2022

This article has been retracted by IOP Publishing following an allegation that raises concerns this article may have been created, manipulated, and/or sold by a commercial entity. In addition, IOP Publishing has seen no evidence that reliable peer review was conducted on this article, despite the clear standards expected of and communicated to conference organisers.

The authors of the article have been given opportunity to present evidence that they were the original and genuine creators of the work, however at the time of publication of this notice, IOP Publishing has not received any response. IOP Publishing has analysed the article and agrees there are enough indicators to cause serious doubts over the legitimacy of the work and agree this article should be retracted. The authors are encouraged to contact IOP Publishing Limited if they have any comments on this retraction.

Retraction published: 23 September 2022

This article has been retracted by IOP Publishing following an allegation that raises concerns this article may have been created, manipulated, and/or sold by a commercial entity. In addition, IOP Publishing has seen no evidence that reliable peer review was conducted on this article, despite the clear standards expected of and communicated to conference organisers.

The authors of the article have been given opportunity to present evidence that they were the original and genuine creators of the work, however at the time of publication of this notice, IOP Publishing has not received any response. IOP Publishing has analysed the article and agrees there are enough indicators to cause serious doubts over the legitimacy of the work and agree this article should be retracted. The authors are encouraged to contact IOP Publishing Limited if they have any comments on this retraction.

Retraction published: 23 September 2022

This article has been retracted by IOP Publishing following an allegation that raises concerns this article may have been created, manipulated, and/or sold by a commercial entity. In addition, IOP Publishing has seen no evidence that reliable peer review was conducted on this article, despite the clear standards expected of and communicated to conference organisers.

The authors of the article have been given opportunity to present evidence that they were the original and genuine creators of the work, however at the time of publication of this notice, IOP Publishing has not received any response. IOP Publishing has analysed the article and agrees there are enough indicators to cause serious doubts over the legitimacy of the work and agree this article should be retracted. The authors are encouraged to contact IOP Publishing Limited if they have any comments on this retraction.

Retraction published: 23 September 2022