Multi-target tracking method for program control simulation equipment teaching training simulation

Abstract

The invention relates to the technical field of image processing, in particular to a multi-target tracking method for program control simulation equipment teaching training simulation, which comprises the following steps of: according to pixel point gray level distribution difference of different area blocks in two adjacent frames of gray level images and pixel point gradient difference between the area blocks, calculating a multi-target tracking result; obtaining a correlation degree between different region blocks in two adjacent frames of gray level images; according to the distribution of pixel points in different area blocks in two adjacent frames of grayscale images, obtaining the position overlapping degree between the different area blocks in the two adjacent frames of grayscale images, and obtaining the distance characteristic value between the different area blocks in the two adjacent frames of grayscale images; and carrying out multi-target tracking according to the distance characteristic value between different region blocks in two adjacent frames of grayscale images. According to the method, the influence of multi-target overlapping is weakened, and the accuracy of multi-target tracking is improved.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a multi-target tracking method for teaching and training simulation of program-controlled simulation equipment. Background Art

[0002] Programmable simulation equipment is often used in teaching, training and experiments to simulate different operating scenarios and conditions. In target tracking training, the simulation equipment needs to use some method to dynamically simulate the movement trajectory of the target and control the equipment to track these targets during training.

[0003] In the conventional multi-target detection and tracking process, multiple targets are tracked by their positional relationship and contours. Since there are multiple targets, there is overlap of multiple targets in the process of tracking multiple targets. Therefore, if multiple targets are tracked only based on their positional relationship and contours, the targets may be lost in the process of tracking, resulting in reduced accuracy of multi-target tracking. Summary of the invention

[0004] The invention provides a multi-target tracking method for teaching and training simulation of program-controlled simulation equipment to solve the existing problems.

[0005] The multi-target tracking method for teaching and training simulation of a program-controlled simulation device of the present invention adopts the following technical scheme: One embodiment of the present invention provides a multi-target tracking method for teaching and training simulation of a program-controlled simulation device, the method comprising the following steps: Acquire a grayscale image of continuous frames; the grayscale image includes a plurality of area blocks; According to the grayscale distribution difference of pixels in different regional blocks in two adjacent grayscale images, the grayscale distribution similarity between different regional blocks in the two adjacent grayscale images is obtained, and the grayscale distribution similarity is corrected according to the pixel gradient difference between the regional blocks to obtain the correlation degree between different regional blocks in the two adjacent grayscale images; According to the distribution of pixel points in different regional blocks in two adjacent grayscale images, the position overlap degree between different regional blocks in two adjacent grayscale images is obtained; according to the position overlap degree between different regional blocks in two adjacent grayscale images, the correlation degree is corrected to obtain the distance characteristic value between different regional blocks in two adjacent grayscale images; Multi-target tracking is performed based on the distance feature values ​​between different area blocks in two adjacent grayscale image frames.

[0006] Furthermore, the specific steps of obtaining the plurality of area blocks contained in the grayscale image are as follows: Each frame of grayscale image is segmented into regions using a semantic segmentation algorithm to obtain several region blocks.

[0007] Furthermore, the grayscale distribution similarity between different regional blocks in two adjacent grayscale image frames is obtained according to the grayscale distribution difference of pixels in different regional blocks in two adjacent grayscale image frames, and the specific steps include the following: Obtain the number of pixels corresponding to each grayscale value in each area block, sort the number of pixels corresponding to all grayscale values ​​in the order of grayscale values ​​0 to 255, and obtain a sequence of the number of pixels in each area block; According to the difference in the mean grayscale values ​​of all pixels in different regional blocks in two adjacent grayscale images and the correlation coefficient between the pixel number sequences in different regional blocks in two adjacent grayscale images, the grayscale distribution similarity between different regional blocks in two adjacent grayscale images is obtained.

[0008] Furthermore, the grayscale distribution similarity between different regional blocks in two adjacent grayscale images is obtained according to the difference in the mean grayscale values ​​of all pixels in different regional blocks in two adjacent grayscale images and the correlation coefficient between the pixel number sequences in different regional blocks in two adjacent grayscale images, and the specific steps include the following: The difference of the mean grayscale values ​​of all pixels in different regional blocks in two adjacent grayscale images is negatively correlated to obtain the first similarity between the two regional blocks; the correlation coefficient between the pixel number sequences of different regional blocks in two adjacent grayscale images is normalized to obtain the second similarity between the two regional blocks; According to the first similarity and the second similarity between the two region blocks, the grayscale distribution similarity between different region blocks in two adjacent frames of grayscale images is obtained.

[0009] Furthermore, the grayscale distribution similarity is corrected according to the pixel gradient difference between the region blocks to obtain the correlation degree between different region blocks in two adjacent grayscale images, which includes the following specific steps: Perform vector sum operation on the gradient vectors of all pixels in each area block to obtain the total gradient vector of each area block; According to the direction difference between the total gradient vectors of different regional blocks in two adjacent grayscale image frames, the grayscale distribution similarity is corrected to obtain the correlation degree between the different regional blocks in the two adjacent grayscale image frames.

[0010] Furthermore, the grayscale distribution similarity is corrected according to the direction difference between the total gradient vectors of different regional blocks in two adjacent grayscale image frames to obtain the correlation degree between different regional blocks in two adjacent grayscale image frames, and the specific steps include the following: The minimum angle between the directions of two total gradient vectors of different area blocks in two adjacent grayscale image frames is recorded as the first angle, and the normalized result of the cosine value of the first angle is recorded as the third similarity; The product result of the third similarity and the grayscale distribution similarity is recorded as the correlation degree between different area blocks in two adjacent grayscale image frames.

[0011] Furthermore, the step of obtaining the position overlap degree between different regional blocks in two adjacent grayscale image frames according to the distribution of pixel points in different regional blocks in two adjacent grayscale image frames includes the following specific steps: Obtain the overlapping area where different area blocks in two adjacent grayscale image frames are mapped to the same coordinate system; The ratio between the number of all pixels in the overlapping area and the number of all pixels in the first area block of different area blocks in two adjacent grayscale images is recorded as a first ratio; the ratio between the number of all pixels in the overlapping area and the number of all pixels in the second area block of different area blocks in two adjacent grayscale images is recorded as a second ratio; The product of the first proportion and the second proportion is recorded as the overlap between different regional blocks in two adjacent grayscale image frames; based on the overlap and the distance between the centroid points of different regional blocks in two adjacent grayscale image frames, the position overlap between different regional blocks in two adjacent grayscale image frames is obtained.

[0012] Furthermore, the position overlap degree between different area blocks in two adjacent grayscale image frames is obtained according to the overlap degree and the distance between the centroid points of different area blocks in two adjacent grayscale image frames, and the specific steps include the following: The product of the overlap degree and the negative correlation mapping value of the distance between the centroid points of different regional blocks in two adjacent grayscale image frames is recorded as the position overlap degree between different regional blocks in two adjacent grayscale image frames.

[0013] Furthermore, the degree of association is corrected according to the degree of position overlap between different area blocks in two adjacent grayscale image frames to obtain the distance feature value between different area blocks in two adjacent grayscale image frames, and the specific steps include the following: The product result of the position overlap degree and the association degree is recorded as a first product value, and negative correlation mapping is performed on the first product value to obtain the distance feature value between different area blocks in two adjacent grayscale image frames.

[0014] Furthermore, the multi-target tracking is performed according to the distance feature values ​​between different area blocks in two adjacent grayscale image frames, and the specific steps include the following: A region block in each grayscale image is used as a node in the directed graph, and all region blocks in each grayscale image are used as a column of nodes in the directed graph according to the centroid point of each region block in each grayscale image in the order from left to right and from top to bottom in the grayscale image; Combine the current frame with the previous adjacent The frame grayscale image is obtained in the directed graph in time order Column nodes, and then use the distance feature values ​​between different area blocks in two adjacent frames of grayscale images as the values ​​of directed edges in the directed graph to obtain a constructed directed graph. According to the Dijkstra algorithm, the shortest path of all nodes in the starting frame of the directed graph is obtained, and multi-target tracking is completed through the shortest path.

[0015] The technical solution of the present invention has the following beneficial effects: the present invention obtains the grayscale distribution similarity between different regional blocks in two adjacent grayscale images based on the grayscale distribution difference of pixels in different regional blocks in two adjacent grayscale images, and improves the accuracy of grayscale analysis between different regional blocks through the grayscale distribution similarity; corrects the grayscale distribution similarity based on the gradient difference of pixels between the regional blocks, and obtains the degree of association between different regional blocks in two adjacent grayscale images, and improves the accuracy of gradient analysis between different regional blocks through the degree of association; corrects the distance feature value between different regional blocks in two adjacent grayscale images based on the distribution and degree of association of pixels in different regional blocks in two adjacent grayscale images, and improves the accuracy of analysis that different regional blocks in two adjacent grayscale images belong to the same target object through the distance feature value; performs multi-target tracking based on the distance feature value between different regional blocks in two adjacent grayscale images to reduce the influence of multi-target overlap and improve the accuracy of multi-target tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 A flowchart of the steps of a multi-target tracking method for teaching and training simulation of a program-controlled simulation device according to the present invention; Figure 2 Flowchart of multi-target tracking for teaching and training simulation of program-controlled simulation equipment. DETAILED DESCRIPTION

[0018] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the multi-target tracking method for teaching and training simulation of a program-controlled simulation device proposed by the present invention, its specific implementation method, structure, features and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0019] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0020] The specific scheme of the multi-target tracking method for teaching and training simulation of program-controlled simulation equipment provided by the present invention is described in detail below with reference to the accompanying drawings.

[0021] See also Figure 1 , which shows a flowchart of a multi-target tracking method for teaching and training simulation of a program-controlled simulation device provided by an embodiment of the present invention, the method comprising the following steps: Step S001: collecting continuous frame images during the teaching and training simulation process of the program-controlled simulation equipment.

[0022] It should be noted that in order to detect and track multiple targets, it is necessary to collect continuous frame images during the training simulation process, and detect and track multiple targets through continuous frame images; since the directly collected images are interfered by noise, filtering processing is required to enhance the image quality.

[0023] Specifically, continuous frame images in the teaching and training simulation process of the program-controlled simulation equipment are collected by a high-speed camera, and the continuous frame images are preprocessed by grayscale conversion and median filtering to obtain preprocessed continuous frame grayscale images.

[0024] Among them, the grayscale and median filtering preprocessing operations are well-known technologies and will not be described in detail here.

[0025] At this point, continuous frame grayscale images of the program-controlled simulation equipment during teaching and training simulation are obtained.

[0026] Step S002: According to the grayscale distribution difference of pixels in different regional blocks in two adjacent grayscale image frames, the grayscale distribution similarity between different regional blocks in two adjacent grayscale image frames is obtained, and the grayscale distribution similarity is corrected according to the pixel gradient difference between the regional blocks to obtain the degree of association between different regional blocks in two adjacent grayscale image frames.

[0027] It should be noted that due to the overlap of multiple targets in the process of detecting and tracking multiple targets, the target may be lost in the process of tracking, resulting in reduced accuracy of multi-target tracking. Therefore, multiple frame images can be selected to analyze the similarity between different regions in all adjacent frame images. The similarity between all adjacent frame regions can be used to use the shortest path in the directed graph to solve the overlap of multiple targets.

[0028] It should be further explained that, since the positions corresponding to different objects in the grayscale image may be relatively different, the regions corresponding to different objects in the grayscale image can be first obtained, and multi-target detection and tracking can be performed through the regions.

[0029] Preferably, each frame of grayscale image is segmented into regions using a semantic segmentation algorithm to obtain a plurality of region blocks in the grayscale image.

[0030] Among them, the semantic segmentation algorithm is a well-known technology and will not be described in detail here.

[0031] The semantic segmentation network used in this embodiment is the DeepLabV3 network, which is a well-known technology, and this embodiment will not repeat the specific results and training methods of the network.

[0032] It should be noted that if the time difference of the same object in consecutive adjacent frames is small, the grayscale difference of the same object is also similar; that is, the grayscale information of the corresponding two regions in adjacent frames still has a high similarity in the grayscale image, that is, the grayscale distribution of pixels in the two regions of the adjacent frames is consistent. Therefore, according to the grayscale distribution difference of pixels in different regional blocks in two adjacent grayscale images, the grayscale distribution similarity between different regional blocks in two adjacent grayscale images is obtained.

[0033] It should be further explained that different area blocks in two adjacent grayscale image frames refer to any area block in each grayscale image frame and any area block in any adjacent grayscale image frame of each frame.

[0034] Specifically, the number of pixels corresponding to each grayscale value in each area block is obtained. When a grayscale value has no corresponding pixel, the number of pixels of the grayscale value is recorded as 0, and the number of pixels corresponding to all grayscale values ​​are sorted in the order of grayscale values ​​0 to 255 to obtain a sequence of the number of pixels of each area block.

[0035] It should be further explained that, since the trend distribution of the pixel number sequence of two different regional blocks is closer, it means that the grayscale distribution of the pixels in the two different regional blocks is similar; and because the difference between the mean values ​​of the grayscale values ​​of all pixels in two different regional blocks is smaller, it can also be explained to a certain extent that the grayscale distribution of the pixels in the two different regional blocks is similar. Therefore, according to the difference in the mean value of the grayscale value of all pixels in different regional blocks in two adjacent frames of grayscale images and the correlation coefficient between the pixel number sequence of different regional blocks in two adjacent frames of grayscale images, the similarity of the grayscale distribution between different regional blocks in two adjacent frames of grayscale images is obtained.

[0036] Preferably, as an embodiment, the specific calculation method of the grayscale distribution similarity between different regional blocks in two adjacent grayscale image frames is: The difference of the mean grayscale values ​​of all pixels in different regional blocks in two adjacent grayscale images is negatively correlated to obtain the first similarity between the two regional blocks; the correlation coefficient between the pixel number sequences of different regional blocks in two adjacent grayscale images is normalized to obtain the second similarity between the two regional blocks; According to the first similarity and the second similarity between the two region blocks, the grayscale distribution similarity between different region blocks in two adjacent frames of grayscale images is obtained.

[0037] In one embodiment of the present invention, it is specifically expressed by the formula: In the formula, Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The correlation coefficient between the pixel number sequences of the area blocks, Indicates The first grayscale image in the frame The mean gray value of all pixels in the area block, Indicates The first grayscale image in the frame The mean gray value of all pixels in the area block, is the absolute value symbol, represents an exponential function with a natural constant as base, for Function, used for normalization; Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The similarity of grayscale distribution between two area blocks. The correlation coefficient between the number of pixel points in two area blocks is calculated by Pearson correlation coefficient, and the process of obtaining Pearson correlation coefficient is a well-known technology, which will not be described in detail here.

[0038] in, It indicates the difference between the mean values ​​of the grayscale values ​​of all pixels in different regional blocks in two adjacent grayscale images. The larger the difference, the less similar the grayscale values ​​between the two regions are, that is, the smaller the similarity of the grayscale distribution is; the smaller the difference, the closer the grayscale values ​​between the two regions are, that is, the greater the similarity of the grayscale distribution is. The larger the correlation coefficient between the pixel number sequences of different regional blocks in two adjacent grayscale images, the greater the similarity of the grayscale distribution between different regional blocks in the two adjacent grayscale images; otherwise, the smaller the similarity of the grayscale distribution between different regional blocks in the two adjacent grayscale images.

[0039] At this point, the similarity of grayscale distribution between different area blocks in two adjacent grayscale image frames is obtained.

[0040] It should be noted that during the teaching and training simulation, there is a certain gradient difference between the area blocks corresponding to each object and the surrounding area blocks. Because the time difference between adjacent frames is small, the gradients between the area blocks corresponding to the same object in two adjacent frames are close. Therefore, the grayscale distribution similarity is corrected according to the pixel gradient difference between the area blocks to obtain the degree of correlation between different area blocks in two adjacent grayscale images.

[0041] Specifically, the gradient value and gradient direction of each pixel in each area block are obtained, and the gradient vector of each pixel is obtained according to the gradient value and gradient direction of each pixel; the gradient value of each pixel is used as the modulus of the gradient vector corresponding to the pixel, and the gradient direction of each pixel is used as the direction of the gradient vector corresponding to the pixel. At this point, the gradient vector of each pixel is obtained.

[0042] A vector sum operation is performed on the gradient vectors of all pixels in each area block to obtain a total gradient vector of each area block. The vector sum operation of vectors is a well-known technique and will not be described in detail here.

[0043] It should be further explained that, since the directional difference between the total gradient vectors of different regional blocks in two adjacent grayscale image frames can reflect the difference between the gradients of the two regional blocks, the grayscale distribution similarity is corrected according to the directional difference between the total gradient vectors of different regional blocks in two adjacent grayscale image frames to obtain the degree of association between different regional blocks in two adjacent grayscale image frames.

[0044] Preferably, as an embodiment, the specific calculation method of the correlation degree between different area blocks in two adjacent grayscale image frames is: The minimum angle between the directions of two total gradient vectors of different area blocks in two adjacent grayscale image frames is recorded as the first angle, and the normalized result of the cosine value of the first angle is recorded as the third similarity; The product result of the third similarity and the grayscale distribution similarity is recorded as the correlation degree between different area blocks in two adjacent grayscale image frames.

[0045] In one embodiment of the present invention, it is specifically expressed by the formula: In the formula, Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The similarity of grayscale distribution between the regional blocks, Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The minimum angle between the directions of the two total gradient vectors of a region block, express The cosine value of Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The degree of association between the regional blocks. Add 1 to the numerator and then divide by 2 to make The range is from 0 to 1.

[0046] Among them, when the angle between the two total gradient vectors of different regional blocks in two adjacent grayscale images is smaller, it means that the gradients between different regional blocks in two adjacent grayscale images are closer, and the degree of association between different regional blocks in two adjacent grayscale images is greater; conversely, the degree of association between different regional blocks in two adjacent grayscale images is smaller. When the grayscale distribution similarity between different regional blocks in two adjacent grayscale images is greater, the degree of association between different regional blocks in two adjacent grayscale images is greater; when the grayscale distribution similarity between different regional blocks in two adjacent grayscale images is smaller, the degree of association between different regional blocks in two adjacent grayscale images is smaller.

[0047] At this point, the correlation degree between different area blocks in two adjacent grayscale image frames is obtained.

[0048] Step S003: According to the distribution of pixel points in different regional blocks in two adjacent grayscale image frames, the position overlap degree between different regional blocks in two adjacent grayscale image frames is obtained; according to the position overlap degree between different regional blocks in two adjacent grayscale image frames, the correlation degree is corrected to obtain the distance feature value between different regional blocks in two adjacent grayscale image frames.

[0049] It should be noted that, since the time interval between two adjacent frames is very short, the position change of the same target is small, the higher the degree of repetition between two regions belonging to the same target in adjacent frames, and the closer the distance between the centroids of two regions belonging to the same target in adjacent frames, the possibility that two regions belong to the same target can be analyzed through the position relationship between different regions in adjacent frames. Therefore, according to the distribution of pixels in different regional blocks in two adjacent grayscale images, the position overlap degree between different regional blocks in two adjacent grayscale images is obtained.

[0050] Preferably, as an embodiment, a specific method for calculating the position overlap degree between different area blocks in two adjacent grayscale image frames is as follows: Obtain the overlapping area where different area blocks in two adjacent grayscale image frames are mapped to the same coordinate system; The ratio between the number of all pixels in the overlapping area and the number of all pixels in the first area block of different area blocks in two adjacent grayscale images is recorded as a first ratio; the ratio between the number of all pixels in the overlapping area and the number of all pixels in the second area block of different area blocks in two adjacent grayscale images is recorded as a second ratio; The product of the first proportion and the second proportion is recorded as the overlap between different area blocks in two adjacent frames of grayscale images; The product of the overlap degree and the negative correlation mapping value of the distance between the centroid points of different regional blocks in two adjacent grayscale image frames is recorded as the position overlap degree between different regional blocks in two adjacent grayscale image frames.

[0051] In one embodiment of the present invention, it is specifically expressed by the formula: In the formula, Indicates The first grayscale image in the frame The number of all pixels in the area block, Indicates The first grayscale image in the frame The number of all pixels in the area block, Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The number of all pixels in the overlapping area between the area blocks, Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The distance between the centroids of the area blocks, represents an exponential function with a natural constant as base, Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The degree of overlap between the locations of the blocks.

[0052] in, It is the ratio of the number of pixels in the overlapping area between different area blocks in two adjacent grayscale images to the number of pixels in the two area blocks. This ratio is used to indicate the relative size of the overlapping area in the two area blocks. The larger the ratio, the larger the overlapping area in the two areas, that is, the larger the corresponding degree of overlap; the smaller the ratio, the smaller the overlapping area in the two areas, that is, the smaller the corresponding degree of overlap. When the distance between the centroids of two different area blocks in two adjacent grayscale images is smaller, it means that the degree of position overlap between the two areas is greater; otherwise, it means that the degree of position overlap between the two areas is smaller.

[0053] Among them, the specific process of obtaining the overlapping area of ​​different area blocks in two adjacent grayscale image frames mapped in the same coordinate system is: taking the lower left corner pixel point on the grayscale image as the coordinate origin, the horizontal right as the horizontal axis, and the numerical upward as the vertical axis, to establish a reference coordinate system; obtaining the overlapping area of ​​different area blocks in two adjacent grayscale image frames in the reference coordinate system.

[0054] At this point, the position overlap degree between different area blocks in two adjacent grayscale image frames is obtained.

[0055] It should be noted that the greater the degree of association between different regional blocks in two adjacent grayscale images, the greater the possibility that the two regional blocks belong to the same target object; the greater the degree of position overlap between different regional blocks in two adjacent grayscale images, the greater the possibility that the two regional blocks belong to the same target object. Therefore, the path distance between different regional blocks in adjacent frames can be obtained by the degree of association and the degree of position overlap, and the optimal path of all targets can be analyzed by the path distance between different regional blocks in adjacent frames to achieve multi-target tracking. Therefore, according to the degree of position overlap between different regional blocks in two adjacent grayscale images, the degree of association is corrected to obtain the distance feature value between different regional blocks in two adjacent grayscale images.

[0056] Preferably, as an embodiment, the specific method for calculating the distance feature value between different area blocks in two adjacent grayscale image frames is: The product result of the position overlap degree and the association degree is recorded as a first product value, and negative correlation mapping is performed on the first product value to obtain the distance feature value between different area blocks in two adjacent grayscale image frames.

[0057] In one embodiment of the present invention, it is specifically expressed by the formula: In the formula, Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The degree of correlation between the regional blocks, Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The degree of overlap between the locations of the blocks. Indicates The first grayscale image in the frame The area block and The first grayscale image in the frame The distance feature value between the area blocks, Represents an exponential function with a natural constant as its base.

[0058] Among them, when the correlation between different regional blocks in two adjacent grayscale images is greater, the distance characteristic value between the two regional blocks should be smaller; otherwise, the distance characteristic value between the two regional blocks should be larger. When the position overlap between different regional blocks in two adjacent grayscale images is greater, the distance characteristic value between the two regional blocks should be smaller; otherwise, the distance characteristic value between the two regional blocks should be larger.

[0059] At this point, the distance feature values ​​between different area blocks in two adjacent grayscale image frames are obtained.

[0060] Step S004: performing multi-target tracking according to the distance feature values ​​between different area blocks in two adjacent grayscale image frames.

[0061] It should be noted that the distance values ​​on different paths can be determined by the distance feature values ​​between different area blocks in two adjacent grayscale images, and the optimal path of all targets can be obtained by analyzing a large number of distance values, so as to complete the tracking of multiple targets.

[0062] Specifically, a parameter is preset , wherein this embodiment is based on This example is described as an example, and this embodiment is not specifically limited. It may depend on the specific implementation situation.

[0063] A region block in each frame of grayscale image is taken as a node in the directed graph, and all region blocks in each frame of grayscale image are taken as a column of nodes in the directed graph according to the centroid point of each region block in each frame of grayscale image in the order from left to right and from top to bottom in the grayscale image.

[0064] Combine the current frame with the previous adjacent The frame grayscale image is obtained in the directed graph in time order Column nodes, and then use the distance feature values ​​between different area blocks in two adjacent frames of grayscale images as the values ​​of directed edges in the directed graph, so as to obtain a constructed directed graph, and obtain the shortest path of all nodes in the starting frame of the directed graph according to the Dijkstra algorithm, and complete the tracking of multiple targets through the shortest path. Among them, the multi-target tracking flow chart of the program-controlled simulation equipment teaching and training simulation is as follows Figure 2 shown.

[0065] Among them, the Dijkstra algorithm is a well-known technology and will not be described in detail here.

[0066] At this point, multi-target tracking is completed.

[0067] It should be noted that the The model is only used to represent negative correlation and constrain the output of the model to be in In the specific implementation, it can be replaced by other models with the same purpose. This embodiment is only based on The model is described as an example without any specific limitation. is the input to the model.

[0068] At this point, this embodiment is completed.

[0069] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-target tracking method for teaching and training simulation of program-controlled simulation equipment, characterized in that: The method comprises the following steps: Acquire a grayscale image of continuous frames; the grayscale image includes a plurality of area blocks; According to the grayscale distribution difference of pixels in different regional blocks in two adjacent grayscale images, the grayscale distribution similarity between different regional blocks in the two adjacent grayscale images is obtained, and the grayscale distribution similarity is corrected according to the pixel gradient difference between the regional blocks to obtain the correlation degree between different regional blocks in the two adjacent grayscale images; According to the distribution of pixel points in different regional blocks in two adjacent grayscale images, the position overlap degree between different regional blocks in two adjacent grayscale images is obtained; according to the position overlap degree between different regional blocks in two adjacent grayscale images, the correlation degree is corrected to obtain the distance characteristic value between different regional blocks in two adjacent grayscale images; According to the distance feature values ​​between different area blocks in two adjacent grayscale images, multi-target tracking is performed; The method of obtaining the similarity of grayscale distribution between different regional blocks in two adjacent grayscale images according to the grayscale distribution difference of pixels in different regional blocks in two adjacent grayscale images comprises the following specific steps: Obtain the number of pixels corresponding to each grayscale value in each area block, sort the number of pixels corresponding to all grayscale values ​​in the order of grayscale values ​​0 to 255, and obtain a sequence of the number of pixels in each area block; According to the difference in the mean grayscale values ​​of all pixels in different regional blocks in two adjacent grayscale images and the correlation coefficient between the pixel number sequences in different regional blocks in two adjacent grayscale images, the similarity of grayscale distribution between different regional blocks in two adjacent grayscale images is obtained; The grayscale distribution similarity is corrected according to the pixel gradient difference between the regional blocks to obtain the correlation degree between different regional blocks in two adjacent grayscale images, and the specific steps include the following: Perform vector sum operation on the gradient vectors of all pixels in each area block to obtain the total gradient vector of each area block; According to the direction difference between the total gradient vectors of different regional blocks in two adjacent grayscale image frames, the grayscale distribution similarity is corrected to obtain the correlation degree between the different regional blocks in the two adjacent grayscale image frames.

2. According to claim 1, a multi-target tracking method for teaching and training simulation of program-controlled simulation equipment is characterized in that: The specific steps of obtaining the plurality of area blocks contained in the grayscale image are as follows: Each frame of grayscale image is segmented into regions using a semantic segmentation algorithm to obtain several region blocks.

3. According to claim 1, a multi-target tracking method for teaching and training simulation of program-controlled simulation equipment is characterized in that: The method of obtaining the similarity of grayscale distribution between different regional blocks in two adjacent grayscale images according to the difference in the mean grayscale values ​​of all pixels in different regional blocks in two adjacent grayscale images and the correlation coefficient between the pixel number sequences in different regional blocks in two adjacent grayscale images includes the following specific steps: The difference of the mean grayscale values ​​of all pixels in different regional blocks in two adjacent grayscale images is negatively correlated to obtain the first similarity between the two regional blocks; the correlation coefficient between the pixel number sequences of different regional blocks in two adjacent grayscale images is normalized to obtain the second similarity between the two regional blocks; According to the first similarity and the second similarity between the two region blocks, the grayscale distribution similarity between different region blocks in two adjacent frames of grayscale images is obtained.

4. According to claim 1, a multi-target tracking method for teaching and training simulation of program-controlled simulation equipment is characterized in that: The grayscale distribution similarity is corrected according to the direction difference between the total gradient vectors of different regional blocks in two adjacent grayscale images to obtain the correlation degree between different regional blocks in two adjacent grayscale images, and the specific steps include the following: The minimum angle between the directions of two total gradient vectors of different area blocks in two adjacent grayscale image frames is recorded as the first angle, and the normalized result of the cosine value of the first angle is recorded as the third similarity; The product result of the third similarity and the grayscale distribution similarity is recorded as the correlation degree between different area blocks in two adjacent grayscale image frames.

5. According to claim 1, a multi-target tracking method for teaching and training simulation of program-controlled simulation equipment is characterized in that: The method of obtaining the position overlap degree between different regional blocks in two adjacent grayscale image frames according to the distribution of pixel points in different regional blocks in two adjacent grayscale image frames includes the following specific steps: Obtain the overlapping area where different area blocks in two adjacent grayscale image frames are mapped to the same coordinate system; The ratio between the number of all pixels in the overlapping area and the number of all pixels in the first area block of different area blocks in two adjacent grayscale images is recorded as a first ratio; the ratio between the number of all pixels in the overlapping area and the number of all pixels in the second area block of different area blocks in two adjacent grayscale images is recorded as a second ratio; The product of the first proportion and the second proportion is recorded as the overlap between different regional blocks in two adjacent grayscale image frames; based on the overlap and the distance between the centroid points of different regional blocks in two adjacent grayscale image frames, the position overlap between different regional blocks in two adjacent grayscale image frames is obtained.

6. A multi-target tracking method for teaching and training simulation of program-controlled simulation equipment according to claim 5, characterized in that: The step of obtaining the position overlap degree between different area blocks in two adjacent grayscale image frames according to the overlap degree and the distance between the centroid points of different area blocks in two adjacent grayscale image frames includes the following specific steps: The product of the overlap degree and the negative correlation mapping value of the distance between the centroid points of different regional blocks in two adjacent grayscale image frames is recorded as the position overlap degree between different regional blocks in two adjacent grayscale image frames.

7. According to claim 1, a multi-target tracking method for teaching and training simulation of program-controlled simulation equipment is characterized in that: The method of correcting the degree of association according to the degree of position overlap between different area blocks in two adjacent grayscale image frames to obtain the distance feature value between different area blocks in two adjacent grayscale image frames includes the following specific steps: The product result of the position overlap degree and the association degree is recorded as a first product value, and negative correlation mapping is performed on the first product value to obtain the distance feature value between different area blocks in two adjacent grayscale image frames.

8. According to claim 1, a multi-target tracking method for teaching and training simulation of program-controlled simulation equipment is characterized in that: The specific steps of performing multi-target tracking according to the distance feature values ​​between different area blocks in two adjacent grayscale image frames are as follows: A region block in each grayscale image is used as a node in the directed graph, and all region blocks in each grayscale image are used as a column of nodes in the directed graph according to the centroid point of each region block in each grayscale image in the order from left to right and from top to bottom in the grayscale image; Combine the current frame with the previous adjacent The frame grayscale image is obtained in the directed graph in time order Column nodes, and then use the distance feature values ​​between different area blocks in two adjacent frames of grayscale images as the values ​​of directed edges in the directed graph to obtain a constructed directed graph. According to the Dijkstra algorithm, the shortest path of all nodes in the starting frame of the directed graph is obtained, and multi-target tracking is completed through the shortest path.

Images