Research on Image Visual Information Coordinate Transformation Model Algorithm

80% of the information obtained by human beings comes from vision, using pictures and videos as information carriers, and the information obtained from them is called visual intelligence. Visual intelligence analysis refers to extracting the size, distance, speed and other parameter information of the target object from the picture, video. In real life, some measuring tools can be used to get the size, distance and speed of the object Equal parameter information, but only through the picture is not directly available Arbitrary The actual distance between the two points needs to establish a transformation model between the plane coordinate system of the image and the actual coordinate system of the image. The mapping between the actual coordinate system and the phase plane coordinate system is established, the mapping relationship between the pixel plane and the pavement coordinate plane and the imaging model of the monocular camera are established. The vanishing point algorithm is used to calculate the distance from the focus to the center of light and the angle between the optical axis and the ground To achieve Based on the picture Picture actual distance between the target objects to facilitate the intelligence analysis of various picture information.


Introduction
Human beings obtain a variety of information to distinguish and treat different things through their own senses as information is ubiquitous, and visual is the main information acquisition methods. The information intelligence obtained from pictures and videos is defined visual intelligence, which refers to information such as the size, distance, and speed of target objects extracted from pictures and videos. Visual intelligence analysis plays a particularly important role in various popular areas such as drone reconnaissance and computer vision [1]. Although researchers have studied some visual systems to obtain relevant information, however it can only be analyzed by a picture or a video under lots of circumstance. The actual distance between the points cannot obtain only through the picture. A transformation model needs to be established between the plane coordinate system of the image and the actual coordinate system, a mapping relationship between the actual coordinate system and the phase plane coordinate system , So that each point in the actual coordinate system can have a corresponding coordinate in the phase plane coordinate system, this conversion relationship is the key to directly obtain the actual distance between the points to be measured through the picture.

Camera Imaging Principle And Ranging Method
The image imaging is related to the parameters inside the camera according to the principle and properties of lens imaging. Different camera parameters determine the optical properties of the picture. Three-dimensional spatial information can be transformed into pictures in two-dimensional space on the basis of the basic principle of a camera which is shown in Figure 1. (2) Figure 1 shows the camera's photographing position under ideal conditions. However, the camera usually has an inclined angle in real life. Formula (2) is used for distance calculation in an ideal state, and it can only calculate the distance from a single point to the camera, and cannot meet the distance measurement under actual conditions. In order to more accurately measure the distance between objects in the picture, we propose a ranging algorithm based on the transformation model of the phase plane and the actual road surface coordinate system.

Ranging Algorithm Based On Phase Plane And Actual Pavement Coordinate System Conversion Model
The distance measuring method [2] of the camera introduced in 2.1 can quickly find the coordinates of each pixel. However, the points in space are arbitrary in practice, and their corresponding relationship isn't a linear relationship. Therefore, it is necessary to establish a transformation model between the phase plane coordinate system formed by the camera and the actual plane coordinate system which can give the position of certain point in phase plane, so the position of this point in actual coordinate system could be figured out. According to the phase plane coordinates of any two points in the picture, the coordinates in the actual coordinate system can be obtained, and then the actual distance of the object can be obtained.
The camera lens is regarded as a pinhole imaging [3], and the transformation of the phase plane coordinate system to the road surface coordinate system requires the transformation of the tilt coordinate system and the camera coordinate system. The mapping relationship is shown in Figure 2.
is the actual road surface coordinate system, and uv is the image plane coordinate system. The four coordinate systems are related to each other. After corresponding mathematical transformation [4], the mapping relationship between the road surface coordinate system and the phase plane coordinate system can be obtained.
The mathematical relationship between the simplified coordinate system uv and the road surface The matrix k in equation (3) is called the internal matrix, and its value is only related to the internal parameters of the camera; the matrix R is called the rotation matrix, and its value is related to the rotation angle t p s 、 、 ; the matrix T is called the translation matrix, which is the phase plane translation to the road surface coordinate system distance. According to equation (3), and the coordinate transformation with any point of the phase plane can be obtain through inverse transformation. Since the road surface is a plane coordinate and has nothing to do with height, let In Equation (4) Assuming that the positions of the two feature points in the phase-plane coordinate system are respectively, the coordinates of the two feature points in the actual road surface coordinate system can be calculated as (3) to (5), then the Actual distance is The transformation model of phase plane to actual road surface coordinates has been established, but three unknown quantities still need to be solved. Assuming the camera's placement position and various parameters are known, the value l t p s 、 、 、 of the rotation angle can be obtained, and three unknown quantities can be solved. Then find the unknown parameters T R k 、 、 by means of the vanishing point method when we can only get the picture data.

Determination Based On Vanishing Point Method
According to the pinhole imaging principle, the camera projects a spatial object onto the projection surface to obtain a single-sided projection image that is closer to the visual effect. Objects close to the camera's origin are shown larger in the picture, and objects farther away are displayed smaller in the picture in the imaging plane, the two actual parallel lines that are not parallel will converge at a point, called the vanishing point [6]. This point is also the projection of an infinite point on the imaging plane, as shown in Figure 3. In Figure 3, the left side is the two parallel lines BD CD AB // AC // 、 in the road surface coordinate system. After shooting by the camera, the straight line AB in the camera coordinate system is no longer parallel to CD , and it will intersect at the point L after extension. Meantime line BC and AD will intersect at point after extension, these two points of intersection are vanishing points.  (7) and (8) show the positional relationship of the three straight lines in the camera coordinate system. Let the vanishing points correspond to the three straight lines N M L 、 、 . In the camera coordinate system, their coordinates are the relationship between the three points It can be expressed by the following equations:  The rotation matrix R can be obtained from equation (10).

Determination of the internal matrix
R . Assuming that the three-point coordinates of L, M, and N are known and the inverse solution is obtained from equation (9), the effective focal length f can be obtained.
According to equation (11), it can be known that the coordinates of two points in a known camera coordinate system can calculate the effective focal length f of the camera, and the value of the internal matrix k can be obtained by being close. Figure 4, the position of the camera's

From equation (12), it is known that
A is the origin of the road surface coordinate system, and the vector OA is the required translation vector. In the process of solving the model, the value OA is solved by knowing the length AB of the reference object, that is, In Equation (13), l is the distance from the origin o of the phase plane to the origin of the actual pavement coordinate system, the matrix T can be derived by passing the points in the selected pavement coordinate system and the corresponding points in the image plane coordinate system.

Determination of the third vanishing point.
In the real state, it is difficult to obtain the third vanishing point directly. The third vanishing point needs to be obtained through mathematical relationships. As shown in Figure 3, the four points in the set ABCD are the four points in the actual road surface, and AB // CD, AD // BC, from the definition and properties of the vanishing point, two parallel straight lines are on the phase plane In the formation of two vanishing points. In Figure 3, L is the vanishing point of the two straight lines AB and CD, and M is the vanishing point of the two straight lines AD and BC. The coordinate system is established by using the lower left corner of the phase plane as the origin. The coordinates of the four points of ABCD are The linear equation is CD : ) ( The linear equation is DA :

M L、
in the camera coordinate system will be

M L、
in the camera coordinate system into formula (9), then the system of equations is From the equation (18), the coordinates of the point N in the phase plane coordinate system and the coordinates in the camera coordinate system can be obtained, thereby deriving the values of the internal matrix k , rotation matrix R , and translation matrix T . Calibration and ranging issues. Therefore, the algorithm flow based on the vanishing point method to solve the transformation model of picture visual information coordinate system is shown in Figure 5.

Solution And Analysis Of Results Based On Phase Plane And Actual Pavement Coordinate Conversion Model
Through data searching [8] , given the static image shown in Figure 6, the distance between the two pillars of the Beijing-style guardrail in the picture is 3.080m, and the parameter of the red vehicle A is 4342 × 1840 × 1500mm. Due to various problems such as shooting equipment and the environment, the quality of the imaging picture is poor, which seriously affects the accuracy of the detection of the target object. It is necessary to first filter the image and filter irrelevant noise to improve the quality of the picture.

Image Preprocessing
In order to reduce the experimental error, Gaussian filtering is performed on the picture to remove the noise of the picture, which facilitates the selection of the feature points of the picture. The filtering result is shown in Figure 7.

Selection Of Feature Points In Image Plane And Solution Of Vanishing Points
It can be known from the properties of the vanishing point that only the coordinates of the two vanishing points in the vertical and parallel directions need to be solved and the actual distance is measured to calculate the corresponding parameters.
The horizontal line and vertical line of the stone brick in the filtered picture shown in Figure 7 are perpendicular to each other, and the calculation mode of the vanishing point is satisfied. Select six points A, B, C, D, E, and F in Figure 7 to calculate the camera parameters, as shown in Figure 8. The actual length of EF the guardrail column is 3.080m. Since BC AD CD AB // // 、 in the pavement coordinate system, the selection of A, B, C, and D meets the conditions for the vanishing point. EF is the selected known actual distance. Using Photoshop software to calibrate the feature points, the coordinates of each point are obtained as shown in Figure 8. The origin of the coordinates is F point, then   Table 1. Project Name The coordinates of the vanishing point in the camera coordinate system are shown in Table 2.

 
According to the rotation matrix, the values of the transformation angles of the three coordinate systems can be derived, as shown in Table 3.

Determine The Actual Distance Of Picture Information Reference Points
It is assumed that the position of the photographer at point B is shown in Figure 9, and the coordinates of A, B, C, and D are marked in Figure 9. That is, the actual distance between the target objects in the picture is calculated based on the picture. The algorithm is useful for extracting various parameter information such as the size, distance, and speed of the target object for various pictures, and then performing intelligence analysis.

4.Conclusion
With the advent of the era of big data, information leakage incidents have emerged endlessly. Not only must we improve our access to effective information, but we must also learn to conceal effective information accordingly to protect ourselves, the motherland, and the homeland. The biggest benefit is visual information information leakage There are a lot of proportions. The article focuses on the