Table of contents

It is our great pleasure to welcome you to the 2021 International Conference on Electronic Communication, Computer Science and Technology (ECCST 2021) on January 07-09, 2022, which is planned to held in Nanchang, China. However, due to the spread of COVID-19, many countries have strict restrictions to travel aboard. Meanwhile, the safety of the participants is always the priority. We made a hard decision to hold the conference online.

2021 International Conference on Electronic Communication, Computer Science and Technology (ECCST 2021) was co-hosted by Jiangxi Science & Technology Normal University. It mainly focused on Electronics and communications, networking, artificial intelligence and computers and other research fields to discuss. This conference provided a valuable opportunity for researchers and industry experts to exchange opinions face-to-face.

ECCST 2021 was a leading conference for about 90 researchers from different countries and territories to present their research results about Electronic Communication, Computer Science and Technology. During the conference, the conference model was divided into three sessions, including oral presentations, keynote speeches, and online Q&A discussion. In the first part, some scholars, whose submissions were selected as the excellent papers, were given about 5-10 minutes to perform their oral presentations one by one. Then in the second part, keynote speakers were each allocated 30-45 minutes to hold their speeches.

List of Committee are available in this pdf.

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

• Type of peer review: Single Anonymous

• Conference submission management system: Morressier

• Number of submissions received: 54

• Number of submissions sent for review: 39

• Number of submissions accepted: 33

• Acceptance Rate (Submissions Accepted / Submissions Received × 100): 61.1

• Average number of reviews per paper: 2

• Total number of reviewers involved: 8

• Contact person for queries:

Name: Xuexia Ye

Email: xx.ye@keoaeic.org

Affiliation: AEIC Academic Exchange Information Centre

The robustness of visual place recognition determines the accuracy of the SLAM to construct the environmental map. However, when the robot moves in the outdoor environment for a long time, it must face the challenge of drastic illumination changes (time shift, season or rain and fog weather factors), which leads to the robot's ability to identify places is greatly restricted. This paper proposes a method for visual place recognition that is more robust to severe illumination changes. First, a generative adversarial network is introduced into visual SLAM to enhance the quality of candidate keyframes, and consistency of geographic locations corresponding to images before and after quality enhancement is evaluated, and then the image descriptor with illumination invariance is established for the robot's new observation image and keyframes of the map. Finally, the performance of the method in this paper is tested on the public dataset. The experimental results show that this method is conducive to improving the quality of environmental map nodes, and enables the robot to show a highly robust ability for visual place recognition in the face of severe illumination changes, which provides a powerful tool for robot to build an accurate environmental map.

This paper presents a low-dropout linear regulator (LDO) with ultra-low power high transient response. This LDO adopts GAP-dynamic-bias circuit to achieve quiescent current as low as 242nA under 0-20mA load. A transient enhancement circuit based on window comparator is used to sense the output voltage change and further enhance the transient response. The circuit is designed and implemented using TSMC0.18um technology, and is verified by simulation using Spectre software. The simulation results show that when the load changes in the range of 10uA-20mA with a 1ns transition edge, the slowest overshoot and undershoot recovery times are only 2.36us and 2.15us. In the input voltage range of 2.5-3.6V, the linear regulation rate and load regulation rate are 0.479mV/V and 2.61nv/mA, respectively.

The traditional BaTiO₃/PbTiO₃ ferroelectric material has excellent performance and is widely used, but the Pb in the material has polluted the environment. Studies have found that other ferroelectric materials can also have good performance after doping and improvement. The problem is that the development is too slow. If a reasonable explanation can be given for the changes from the original structure, the research and development process will be accelerated. In this paper, we investigated the ferroelectric properties at the orthogonal-crystal KNbO₃ using first-principles calculations coupled with a phenomenological theory model. By replacing the K atoms with Na, studying the stability, polarization, energy band, and density of states after replacement doping, explaining the internal mechanism of its generation, and providing theoretical support for the research and development of new materials.

Because of using correlation processing technology, it is difficult for traditional jamming patterns to achieve the desired jamming effect on phase coded radar. At present, the forward deception jamming of phase coded radar is mainly realized by uniform interrupted sampling repeater jamming (UISRJ). In order to solve the problem that the UISRJ can only produce a single and delayed false target, a kind of non-uniform interrupted sampling repeater jamming (NUISRJ) is proposed. The non-uniform sampling pulse train is generated through pseudo random sequence, and then sampled and forwarded to generate jamming signal, which is processed by radar pulse compression to form dense false targets. Combining the theoretical analysis and simulation results, it can be seen that the NUISRJ has the dual effects of suppressing jamming and spoofing jamming, and the distribution of false target string is more random, including the leading and lagging false target. The amplitude and coverage area of false targets group can be controlled by adjusting sampling pulse width and forwarding times to effectively implement jamming.

Satellite channel is the representative non-linear channel with limited frequency band. The random change of the channel causes the communication stability to be affected. With the development of satellite communication technology, the bandwidth of satellite communication has increased and the data rate has also increased significantly, resulting in more serious code interference, which has become a key factor affecting communication quality. Finding an efficient equalization technology to eliminate channel inter-code interference is the key to ensuring satellite broadband communications. This article first briefly discusses the principle of eliminating inter-symbol interference and common equalizer algorithms, and summarizes the main factors that affect the performance of channel equalization algorithms. On this basis, an improved LSTM blind equalization algorithm is proposed.

Measuring the orbital angular momentum(OAM) mode of vortex beams is of great significance in applications based on vortex beams.we propose a method of phase matching characteristics of vortex beams to measure the OAM mode. The method first use a high-speed spatial light modulator(SLM) to sequentially load a set of helical phase sequence images, so that the vortex beam is modulated by this SLM. Then obtain the modulated optical field through pinhole filtering, and then the synchronized tilt phase modulation is performed by the 4f system and high-speed SLM. When the OAM mode of vortex beam is opposite to the topological charge of helical phase image, that is, the incident vortex beam degenerates to the fundamental Gaussian mode beam,and through our optical system, the output plane obtained a indicating spot with the position related to the OAM mode. The Simulation and experimental results show that the method obtains a clear OAM mode indicating spot on the final output plane, which verifies the theoretical derivation.

In response to the needs of spacecraft cable three-dimensional crossover layout, research on cable installation methods based on CSB bus has been carried out. Aiming at the status of using CSB bus cable in the satellite communication cabin of DFH-4 platform for the first time, on the basis of detailed analysis of the performance of CSB bus, the laying and binding methods are analyzed and tested, and the cable installation method on the carbon fiber bracket is proposed. Technical solutions. It has been proved through practice that this method meets the actual needs of satellites, solves the problems of laying and tying the bus cables of the communication cabin, and effectively meets the needs of the spacecraft cable three-dimensional crossover layout, and greatly improves the spacecraft assembly efficiency.

In this work, an implantable loop antenna with Hilbert curve structures is proposed. This antenna uses Hilbert curves to achieve extremely small electrical dimensions. In addition, we use magnesium instead of copper in the antenna structure, as it is healthier than copper. The proposed antenna covers a frequency band ranging from 765 to 1143 MHz (S11< -10 dB), a high efficiency (1.5%), and a maximum realized gain of -15.42 dBi. It is a promising candidate for implantable capsule antenna in medical devices.

With the progress of society, more and more families take cars as the main means of transportation. At the same time, due to the negligence of some parents, there are more and more suffocation accidents caused by children left in the car and adults resting in the car. At present, the research can only detect people who are moving around in the car, but not those who are resting in the car. At present, many systems are detecting and alerting, rather than directly solving the problem. The people left in the car with engine cut off and all the doors closed.The capacity of the car battery is very small, which requires the system operates with minimal power consumption. This problem is not considered in the current relevant research. To solve the above problems, this paper designs a system based on low power consumption dual microprocessor, combined with power management program, equipped with carbon dioxide sensor, temperature sensor and camera, which has low power consumption and can detect people resting and staying in the car at the same time. The system can effectively solve the shortcomings of the current research in view of the two situations of people resting and staying in the car at the same time.

On the one hand, target tracking technology is widely used in navigation, guidance, alarm and monitoring and other military fields, on the other hand, it is also widely used in civil fields such as traffic information monitoring and security. This paper mainly studies the application of information processing system based on DSP in electronic information engineering. This paper introduces DSP information processing system in detail. Aiming at the comprehensive requirements of real-time and tracking performance of the target tracking system, multi-feature fusion optimization method is adopted to improve the particle filter tracking algorithm to improve the accuracy of the tracking algorithm. In this paper, the simulation results show that the improved algorithm has a certain scale and occlusion robustness, and can adapt to long-term stable real-time tracking.

Aiming at the problem that civil aviation radio navigation equipment is susceptible to interference more and more frequently, through the study of aviation radio interference theory, the categories of civil aviation radio interference are classified, and then the source and harm of radio interference are analyzed, and specific measures to solve the interference problem are proposed. This is of great help to solve the difficult problem of radio interference signals, and is conducive to reducing the probability of civil aviation radio navigation equipment being interfered, thereby ensuring aviation flight safety.

Through the operation and maintenance system, the centralized control of the operation state of enterprise equipment can be realized. In order to solve the problem of unstable monitoring effect of traditional methods, untimely risk early warning and high error rate of early warning information, a lightweight it operation and maintenance integrated monitoring method for APP system is proposed. Analyze the actual needs of the monitoring function of the lightweight IT operation and maintenance system, design a targeted IT operation and maintenance monitoring function module according to the analysis results, and give the lightweight IT operation and maintenance process. A lightweight IT operation and maintenance database is established based on fuzzy theory, and data with similar characteristics is simulated to minimize the dispersion of sample data. Finally, in order to achieve integrated monitoring of lightweight IT operation and maintenance, and to improve monitoring efficiency and monitoring accuracy, it is necessary to reduce the dimensionality of multi-dimensional data, which is mainly achieved through principal component analysis. The experimental results show that compared with the traditional method, the method in this paper can obtain accurate early warning information in time in monitoring, indicating that its monitoring effect is better, and it has a certain positive effect on ensuring the safe operation of the IT operation and maintenance system.

Object detection is a challenging task in computer vision, which aims to classify and locate objects. SSD (Single Shot MultiBox Detector) is one of the classical algorithms in the current object detection field, which uses multiple feature maps with different scales to detect all objects in an image. Based on SSD, in this paper, we propose an improved AFESSD for object detection in natural gas pipeline construction scene: Firstly, the attention mechanism module is introduced; Secondly, Feature Fusion Block and Feature Enhancement Block are designed to achieve high-to-low level feature information fusion and enhancement. Finally, we verify AFESSD on two object detection datasets of natural gas pipeline construction scene. Experimental results on the first dataset demonstrate that SSD and AFESSD achieve 81.35% and 83.13% mean average precision (mAP) respectively. On the second dataset, SSD and AFESSD achieve 50.55% and 50.99% mAP respectively. Therefore, AFESSD improves SSD's object detection accuracy in natural gas pipeline construction scene.

Artificial intelligence has been widely utilized in people's lives, especially in communication. The article provides a review of the recent research status of the application of artificial intelligence in communication, the usefulness of big data, machine learning, and cloud computation in this field. Several examples of how artificial intelligence is used to communicate are illustrated to discuss better and form a framework of modern artificial intelligence applications. The aspects of concern in ethics of using artificial intelligence are also mentioned, as ethics are the principles judging whether the technologies are suitable for humans to use or not. In order to predict the future development of communication, an introduction of previous progress in the field of communication is provided and sorted into categories of wireless communication, ethical concerns, and network monitor and control. This study confirms that the future development of artificial intelligence is mainly in two aspects: SDN (Software-defined network) and NFV (Network function visualization). AI will thus be used in enhancing communication efficiency.

On the basis of completing the division of geographical regions, for the problem of satellite mobile handover, this paper proposes a satellite selection algorithm for selecting the satellite with the longest coverage time to reduce the handover rate. Also based on the idea of "centralized location management", this paper proposes the satellite handover scheme based on the geographical division scenario to update the satellite routing table entries and ensure the correct routing of data packets. Regarding the user handover problem, based on the idea of "decoupling data transmission and address update", a user handover scheme based on geographical division scenario is proposed, which reduces handover delay and signaling overhead compared to traditional satellite network handover schemes.

The structure, electronic and ferroelectric properties of BSZT were investigated based on first-principles calculations. The result of relaxation structure and two-dimensional charge density indicates that the tetragonal crystalline structure is distorted after Sr and Zr doping. The double-well curve fits well with the phenomenological Landau-Devonshire theory. Moreover, the smaller Zr ion displacement leads to a lower spontaneous polarization of BSZT than the original BaTiO₃. The hybridization between the 3d states of Ti and the 2p states of O, and the hybridization between the 4d states of Zr and the 2p states of O are observed in the density of states, which is the reason for the ferroelectricity of BSZT materials.

Due to the development of communication technology, mobile devices continue to increase. As one of the critical technologies of 5G, D2D communication is an up-and-coming technology. In this paper, multiple D2D pairs usually multiplex the same channel, which will cause serious channel interference. To solve this problem, a distributed deep reinforcement learning framework is proposed, which can well adapt to the power allocation in a dynamic environment. The simulation results show that compared with other benchmark methods, the proposed scheme can improve the overall D2D rate.

With the miniaturization of RF circuits, the transmission of signals in different layers becomes important. How to realize the vertical transition structure with wide transmission band and low loss has become the focus of research. In this paper, we proposed a broadband vertical transition structure design based on cavity coupling. This model achieves broadband coupling between the upper and lower microstrips by adding parabolic patches, which improves the model transmission bandwidth. The simulation results show that the optimized model has ultra-wideband transmission characteristic, the 1dB bandwidth is 3.32 GHz-10.42 GHz and the fractional bandwidth is more than 103%. What' more, the thick ground layer of the model improves the heat dissipation performance. Therefore, this structure is very suitable for high-speed and high-density RF circuits.

Compared with the performance index of UAV system antenna, a compatible trigeminal monopole array antenna is proposed. This antenna uses photosensitive resin as the substrate material, and can be made with UAV shell using 3D printing technology, realizing the integration of the antenna system and UAV. The antenna uses a trigeminal monopole antenna as the basic radiation unit, which is arranged in accordance with the Sierpinski fractal structure to form fractal radiation array. While enhancing the radiation through the array, it ensures that the antenna has a larger working bandwidth. In radiation performance test, the antenna is operating over a frequency spectrum of 1.038-11.916 GHz, and the minimum S11 is -37.92 dB. The proposed antenna has multi-band compatibility, better radiation performance, and greater performance redundancy.

In the past ten years, deep learning has achieved remarkable results in the area of natural image segmentation, and has gradually turned to the field of medical image segmentation. The precise segmentation of spine images can be used for early screening of spondylopathy, which is convenient for early detection and treatment of patients. Aiming at the segmentation of the spine image by the U-Net network, which structure will lead to large model calculation, network overfitting, image size, noise information and other issues. This paper introduces a new method based on spatial pyramid pooling module ASpp on U-Net network and applies the proposed network structure to the segmentation of the spine image. The model is first enhanced by the Spp module and Densenet data; secondly, the U-Net encoding and decoding structure is adopted, and the deep separable convolution is used. This method greatly reduces the complexity and computation of the model, and uses a rectangular convolution kernel to increase the computation of the model in a small amount. Based on the volume, the receptive field of the convolution operation is expanded; finally, in order to effectively enhance the segmentation area of the image, the SE-Net attention mechanism module is added to the lateral connection. The method proposed in this paper conducts ablation experiments on a public dataset of spine images and compares it with the U-Net network. The method proposed IOU in this paper can reach 72%, and the F1-score can reach 0.92. The comparison of the experimental results of spine medical image datasets shows that the method proposed in this paper can effectively improve the accuracy and accuracy of spine image segmentation.

Aiming at the problem of the detecting difficulty and imaging the moving target submerged in clutter, a multi-channel Synthetic Aperture Radar (SAR) moving target detection and imaging method based on complex Convolutional Neural Network (CNN) is proposed. First, using Range Doppler algorithm (RDA) to image SAR data to obtain images with clutter, noise and defocused moving targets. Then, the SAR image is used as the input of complex CNN. A SAR image with the clutter and noise eliminated, and the moving target in focus is obtained after processing by neural network. The simulation experiment results show this method is effectiveness.

Quantum NOT gates play an important role in the process of quantum information conversion. However, when the X-gate operation is executed on a real quantum computer, there is a large deviation between the actual operation result and the theory, which will lead to inaccurate results when the quantum algorithm containing the X-gate operation is executed. In order to facilitate users to understand the error fluctuations of the X gate in time before executing the quantum algorithm containing the X gate operation on the IBM quantum cloud platform, this paper proposes a method to measure the X gate error. By measuring the X-gate error of four small-scale superconducting quantum computers on the back end of the IBM quantum cloud platform, analyze the degree of fluctuation of the actual measurement value of the quantum device; at the same time, under different execution times shots, the influence of the X-gate test is analyzed. The test results show that the measured value of each qubit of different quantum devices fluctuates to different degrees; the execution times of shots are different, and the test results of X-gate error will be affected to different degrees. This test method helps users select the optimal performance of the qubits before executing quantum algorithms with X gate on IBM's small-scale quantum computers.

In the field of small object detection, Yolov4-Tiny is inadequate in feature extraction and does not make best of multi-scale features. In this paper, an improved BiFPN framework is proposed based on Yolov4-Tiny to increase object detection precision. Moreover, the Yolov4-Tiny is taken as the backbone network and introduce spatial pyramid pooling (SPP) to connect and merge multi-scale regions. Finally, our method can achieve 79.53% map on Pascal VOC dataset, which is 2.12% higher than the original Yolov4-Tiny model.

Classifying subcategories of the same category (such as birds, cars, airplanes) is mainly to find discriminative features and accuracy regional positions in the fine-grained visual classification (FGVC). In this article, we propose to use global average pooling slice feature maps to find significant discriminative regions without complicated network designs or training operations and use Drop-Block mechanisms to solve the problem of network overfitting. Specially, we use the feature maps of the multi-branches as inputs, and average-pooling them with different sizes of convolution kernels to obtain feature maps that containing deeper and shallow information for getting more accuracy granularities. Our methods are called as More Accurate Multi-Granular Convolutional Neural Network (MAG-CNN). Compared with other networks with more complex designs, the network only requires common operations such as pooling and convolution to achieve higher accuracy. The MAG-CNN can be trained end-to-end, without any bounding-box, and our method has reached the most advanced performance on three common fine-grained image classification benchmark datasets (CUB-Birds, FGVC-Aircraft, and Stanford-Cars).

Gyrowheel is a new electromechanical servo device which can realize the integration of spacecraft attitude control and attitude measurement. Due to the existence of machining and assembly errors, there must be a deviation between the rotor centroid of the gyrowheel and the support center, that is, the centroid offset. Firstly, to solve this problem, the dynamic modeling of rotor center of mass offset is carried out from the perspective of vector mechanics by using D'Alembert principle. Secondly, the established dynamic model is simplified, and the influence of rotor center of mass offset on the micro vibration of gyrowheel is analyzed. Finally, the correctness of this method is verified by simulation.

The rapid development of the Internet has pushed society into the era of information explosion, and people are faced with more and more information screening and choices. A recommendation system is an effective way to process massive amounts of information, and it is also a tool that can make recommendations based on user behavior. Traditional collaborative filtering algorithms generally use the cosine similarity formula to calculate the similarity between users or items to make recommendations. Due to the popularity of the Internet, more and more popular items have appeared. The appearance of popular items not only affects the recommendation results, but also fails to reflect the real needs of users. This paper proposes an improved formula for cosine similarity with a penalty factor, which can restrain the influence of popular items on the recommendation result. Finally, the Movie Lens data set is used to verify that the recommended performance indicators have been improved to a certain extent.

A data enhancement method for object detection is proposed to address the problem that there are few data enhancement means for target detection. This method has obvious advantages for small targets and scenarios where the target background is lacking. It is also a method to solve the problem of small and unbalanced samples. In this paper, the object is subtracted from the bounding box, and after processing, it is fused into a new image using various image classification improvement means such as random brightness, random contrast, random rotation, random cropping, etc. Using mixup-like fusion methods, and XIOU is proposed to guarantee the t intersection ratio of the target synthesis during synthesis. An improvement of 2% is obtained on yolov3 over the VOC dataset.

Accurate face detection and tracking is widely used in many social life scenes. However, the uncontrollable background noise and random illumination change in the application scene will reduce the detection accuracy of the tracked target, and the rotation, occlusion and overlap of the tracked target will also affect the accuracy and success rate of the tracking algorithm. In order to solve the above problems, this paper proposes a method that uses deep learning detection method to supervise the correlation filter tracking algorithm to improve the success rate of de-tection and tracking. First of all, in the first frame of the picture, we use the deep learning SSD (Single Shot MultiBox Detector) algorithm to detect the face, and take the detected face as the tracking target, and use the correlation filtering algorithm DSST (Discriminative Scale Space Tracker) Tracking in the process of tracking, face detection is continuously carried out on the tracking target, and the detection results are used to monitor the tracking results, so as to reduce the target drift caused by the boundary effect, thus improving the accuracy of the tracking algorithm. The algorithm is tested and verified on OTB100 data set. The final experimental results show that the accuracy of this algorithm is obviously better than the mainstream classical algorithm, and the frame rate meets the real-time re-quirements.

The energy consumption of traditional mobile network is mainly reflected by task calculation and data transmission. The edge server in mobile edge computing is placed at the edge of the network closer to the mobile terminal device. When the computing resources of the mobile terminal device are insufficient, the task can be offloaded to the edge server for processing. This paper discusses the energy consumption optimization in the Mobile Edge Computing (MEC) system, which is composed of the edge server and mobile devices at the edge of the network. Firstly, the system model of MEC is elaborated. Secondly, the energy consumption optimization for mobile devices and base stations is studied. Which oriented to the mobile device energy consumption optimization is divided into multiple users - single MEC server scenarios of energy consumption optimization and multiple users - the energy optimization of MEC server scenario, and enumerate the oriented to the base station and a new method for the optimization of energy consumption for mobile equipment energy consumption optimization method, in addition to the insufficiencies of these methods and the prospect of the future are also explained.

Aiming at the problems of long time-consuming monitoring and poor monitoring accuracy in traditional customer service data analysis platform operation abnormality monitoring methods, a customer service data analysis platform operation abnormality monitoring method based on the improved FP-Growth algorithm is designed. Obtain customer service data sets, classify data types, filter customer behavior, identify the operating status of the data analysis platform, improve the FP-Growth algorithm to build a rule configuration model, set the platform safety factor threshold, and keep the reconstruction error of customer service data to a minimum Within the scope, optimize the abnormal monitoring mode. The experimental results show that the average recovery time of the proposed customer service data analysis platform operation anomaly monitoring method is 5.239s, and the average platform operation anomaly accuracy rate is 97.3%, indicating that the customer service data analysis platform integrated with the improved FP-Growth algorithm operates abnormally The monitoring method performs better.

Quantum secret sharing(QSS) is a cryptographic protocol based on the uncertainty principle and quantum no-cloning theorem and is an essential aspect in quantum computing. This paper describes the Grover search algorithm and how to improve QSS protocol by Grover Algorithm. The QSS based on the Grover algorithm is more secure and easier to operate compared to old QSS protocols. Some scientists improved the security of the first QSS protocol based on QSA by changing the check mode. After that, a protocol with quite different procedures was proposed. The improved one enhances the feasibility of the protocol and it could also make sure that the communication is secure from the attacks.

This article briefly introduces the principles and advantages of optical fiber transmission and the characteristics of the integrated IP core developed by Xilinx. Aiming at the advantages of optical fiber communication, Xilinx ZYNQ7000 series FPGA chips are used to design a high-speed data optical fiber transmission scheme based on FPGA. This solution uses the method of creating Block Design for design and development. It uses Verilog hardware description language and Vivado development software for experimental verification. In the FPGA simulation environment, the code is converted into a binary file and written to the chip through the JTAG interface for testing, and the experimental data is analyzed. As a result, it meets the design requirements of high-speed data transmission, which is of great significance to the research of high-speed data optical fiber transmission.

Aiming at the problem that Yolov5 is difficult to accurately detect backlighting and small samples in national grid power transmission and distribution operations, studies the illegal behavior technology of "two wear and one wear" in power grid operation, and proposes a violation recognition algorithm based on the enhanced YOLOv5. First, add a new detection layer and use the BiFPN (bi-directional feature pyramid network) layer for feature fusion, so that feature layers of different scales can better learn the weight distribution and enhance the fusion ability. Secondly, add CBAM (Convolutional Block Attention Module) module before the output detection layer feature map to make full use of channel and spatial information to achieve better model accuracy and recall. The experimental results on the six self-made data sets show that the mAP of this method is 91.0%, which is 5.6% higher than the original algorithm, indicating that the model has stronger predictive ability and robustness for the identification of personnel and safety tools in transmission and distribution scenarios.