Long jump evaluation method and device, electronic equipment and storage medium

Abstract

The application provides a long jump evaluation method and device, electronic equipment and a storage medium, wherein the method comprises: determining a to-be-tested video under a long jump evaluation area and a scale line coordinate; determining a landing frame from each frame in the to-be-tested video; performing foot segmentation on the landing frame based on the coordinates of the foot skeleton points of the personnel in the landing frame to obtain the coordinates of a plurality of foot contour points in the landing frame; and performing long jump result evaluation based on the coordinates of the plurality of foot contour points in the landing frame and the scale line coordinate, thereby overcoming the defect of inaccurate long jump result calculation in the traditional scheme, analyzing the state of the test personnel through human body detection, posture estimation and foot segmentation analysis to determine the landing frame corresponding to the landing frame, and performing result evaluation accordingly, thereby improving the accuracy of long jump result calculation; in addition, the use of computer vision technology and a cycle state machine for long jump evaluation also improves the long jump evaluation efficiency and the teaching efficiency of the physical education class.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, and in particular to a method, apparatus, electronic device, and storage medium for long jump evaluation. Background Technology

[0002] In recent years, sports have received increasing attention. This is evident in the rising proportion of physical education scores in the high school entrance examination, demonstrating the growing importance of physical education in primary, junior high, and higher education. The long jump, a classic track and field event, is a staple of both compulsory and higher education.

[0003] Current long jump evaluation processes largely rely on manual methods or infrared measurement equipment. This involves measurement by human ergometers or using infrared measuring devices. However, manual measurement is highly subjective; different personnel may render different judgments when judging the same test subject due to subjective differences. Infrared measurement requires complex and cumbersome equipment, making the process inconvenient and contradicting the principle of simplicity and ease of use in physical education. Furthermore, current score calculations often use the shortest distance between the test subject's foot outline and the take-off line in the landing image, but camera distortion can lead to inaccurate long jump results. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, and storage medium for long jump evaluation, in order to solve the defects of inaccurate performance calculation in the prior art.

[0005] This invention provides a long jump evaluation method, comprising:

[0006] Determine the coordinates of the video to be tested and the scale lines within the long jump evaluation area;

[0007] Determine the landing frame from each frame in the video to be tested;

[0008] Based on the coordinates of the foot bone points of the person in the landing frame, the landing frame is segmented into foot parts to obtain the coordinates of multiple foot contour points in the landing frame.

[0009] The long jump performance is evaluated based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines.

[0010] According to a long jump evaluation method provided by the present invention, the step of evaluating the long jump performance based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale line includes:

[0011] Based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale line, the scale interval where the multiple foot contour points are located is determined.

[0012] The person's long jump score is determined based on the upper or lower limit of the scale interval and the distance between the plurality of foot contour points and the upper or lower limit of the scale interval.

[0013] According to a long jump evaluation method provided by the present invention, determining the scale interval where the multiple foot contour points are located based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale line includes:

[0014] Curve fitting is performed on multiple foot contour points in the landing frame, and the foot contour curves obtained from the curve fitting are then sparsified.

[0015] Based on the coordinates of multiple foot contour points after sparsification, and the coordinates of the scale line, the scale interval in which the multiple foot contour points are located is determined.

[0016] According to the long jump evaluation method provided by the present invention, the coordinates of the foot bone points are determined based on the following steps:

[0017] Based on the attitude estimation model, the attitude of the person in the landing frame is estimated to obtain the coordinates of the foot bone points of the person in the landing frame;

[0018] The posture estimation model is trained based on a first sample image, the region label of the long jump evaluation area in the first sample image, and the human skeleton point label, with the constraint that the person is in the long jump evaluation area for long jump evaluation.

[0019] According to a long jump evaluation method provided by the present invention, determining the landing frame from each frame of the video to be tested includes:

[0020] Based on the human detection result of the first frame in the video to be tested, the state of the person is determined, and when the state is in the ready state, the first movement speed of the foot bone points of the person is obtained based on each frame in the video to be tested.

[0021] When the first moving speed indicates that the person is in a jumping state, the second moving speed of the person's foot bone points is obtained based on each frame in the video to be tested.

[0022] If the second moving speed is less than a preset landing speed threshold, it is determined that the person is in a landing state, and the landing frame corresponding to the landing state is determined from each frame in the video to be tested.

[0023] According to a long jump evaluation method provided by the present invention, the step of obtaining the second movement speed of the foot bone points of the person based on each frame in the video to be tested includes:

[0024] From each frame in the video to be tested, determine the jump frame corresponding to the jump state;

[0025] Based on the coordinates of the person's foot bones in the take-off frame and the coordinates of the person's foot bones in the second frame after the take-off frame in the video to be tested, the displacement of the person's foot bones is determined.

[0026] Based on the coordinates of the person's foot bones in the second frame, the displacement of the person's foot bones, and the coordinates of the take-off line in the long jump evaluation area, the take-off state is verified, and if the verification passes, the second movement speed of the person's foot bones is obtained based on each frame after the second frame in the video to be tested.

[0027] According to a long jump evaluation method provided by the present invention, the take-off state is determined based on the following steps:

[0028] Based on the first moving speed, determine the horizontal moving speed component and the vertical moving speed component of the person's foot bone points.

[0029] If the horizontal velocity component and the vertical velocity component are both greater than a preset soft velocity threshold, and the horizontal velocity component or the vertical velocity component is greater than a preset hard velocity threshold, then the person is determined to be in a jumping state.

[0030] The preset soft speed threshold is less than the preset hard speed threshold.

[0031] The long jump evaluation method provided by the present invention further includes:

[0032] When the state is in the ready state, pose estimation is performed on the first frame to obtain the coordinates of the foot bone points of the person in the first frame;

[0033] Based on the coordinates of the foot bone points of the person in the first frame, the first frame is segmented into foot parts to obtain the coordinates of multiple foot contour points in the first frame.

[0034] Based on the coordinates of multiple foot contour points in the first frame and the coordinates of the take-off line under the long jump evaluation area, a violation assessment of stepping on the line in the long jump is performed.

[0035] The long jump evaluation method provided by the present invention further includes:

[0036] Based on the jump frame in the video to be tested, and the third frame before the jump frame, the third movement speed of the foot bone points of the person's left and right feet is obtained.

[0037] The long jump take-off violation is assessed based on the third movement speed of the foot bones of the person's left and right feet, and a preset violation speed threshold.

[0038] According to a long jump evaluation method provided by the present invention, at least one of the preset landing speed threshold, the preset soft speed threshold, and the preset hard speed threshold is determined based on the coordinates of the scale line under the long jump evaluation area.

[0039] The present invention also provides a long jump evaluation device, comprising:

[0040] The determination unit is used to determine the coordinates of the video to be tested and the scale lines within the long jump evaluation area;

[0041] The landing frame determination unit is used to determine the landing frame from each frame in the video to be tested;

[0042] The contour point determination unit is used to segment the landing frame into foot sections based on the coordinates of the foot bone points of the person in the landing frame, and obtain the coordinates of multiple foot contour points in the landing frame.

[0043] The performance evaluation unit is used to evaluate the long jump performance based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale line.

[0044] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the long jump evaluation method as described above.

[0045] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the long jump evaluation method as described above.

[0046] The long jump evaluation method, apparatus, electronic device, and storage medium provided by this invention determine the test video and scale line coordinates within the long jump evaluation area. From each frame of the test video, the landing frame is determined. Based on the coordinates of the foot bone points of the person in the landing frame, the landing frame is segmented to obtain the coordinates of multiple foot contour points in the landing frame. Based on the coordinates of these multiple foot contour points and the scale line coordinates, the long jump performance is evaluated. This overcomes the inaccuracy of traditional long jump performance calculations. By analyzing the state of the tester through human detection, posture estimation, and foot segmentation, the corresponding landing frame is determined, and the performance is evaluated accordingly, improving the accuracy of long jump performance calculation. Furthermore, the use of computer vision technology and a cyclic state machine for long jump evaluation also improves the efficiency of long jump evaluation and the teaching efficiency of physical education classes. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating the long jump evaluation method provided by the present invention;

[0049] Figure 2 This is a schematic diagram of the scale distribution under the long jump evaluation area provided by the present invention;

[0050] Figure 3 This is a flowchart illustrating step 140 in the long jump evaluation method provided by the present invention;

[0051] Figure 4 This is an example diagram illustrating the evaluation process for long jump performance provided by the present invention;

[0052] Figure 5 This is a flowchart illustrating step 141 of the long jump evaluation method provided by the present invention;

[0053] Figure 6 This is a flowchart illustrating step 130 in the long jump evaluation method provided by the present invention;

[0054] Figure 7 This is a flowchart illustrating step 120 in the long jump evaluation method provided by the present invention;

[0055] Figure 8 This is a schematic diagram of the cyclic state machine in the long jump evaluation method provided by the present invention;

[0056] Figure 9This is a flowchart illustrating step 122 in the long jump evaluation method provided by the present invention;

[0057] Figure 10 This is a schematic diagram of the process for determining the take-off state provided by the present invention;

[0058] Figure 11 This is a schematic diagram of the long jump violation assessment process provided by the present invention;

[0059] Figure 12 This is a schematic diagram of the long jump take-off violation assessment process provided by the present invention;

[0060] Figure 13 This is a general framework diagram of the long jump evaluation method provided by the present invention;

[0061] Figure 14 This is a schematic diagram of the structure of the long jump evaluation device provided by the present invention;

[0062] Figure 15 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0064] In recent years, sports have received increasing attention. This is evident in the rising proportion of physical education scores in the high school entrance examination, demonstrating the growing importance of physical education in primary, junior high, and higher education. The long jump, a classic track and field event, is a staple of both compulsory and higher education.

[0065] The current long jump evaluation process is relatively traditional, mostly conducted manually or with the aid of infrared measuring equipment. However, manual measurement is prone to different judgments due to the subjective differences of the measuring personnel; while infrared measurement requires the deployment of complex and cumbersome testing equipment, which contradicts the principle that physical education courses should be simple and easy to use.

[0066] The following points need to be considered in the traditional long jump evaluation process:

[0067] First, performance evaluation, which requires providing the tester's long jump score;

[0068] Secondly, violation detection, which requires determining whether the tester violated any rules during the testing process and, if so, identifying the type of violation.

[0069] Thirdly, key postures, which means capturing key frames of the tester's actions during the test, such as preparation, jumping, airborne, and landing;

[0070] Fourth, process analysis, which requires providing several indicators of the testers during the test, such as arm swing angle, knee flexion angle, take-off angle, height in the air, and trajectory in the air.

[0071] In evaluating performance, the landing frame is often used as the basis for the long jump score. This means that the shortest distance between the outline of the test taker's feet and the take-off line in the landing frame is used as the long jump score. However, if there is distortion in the camera, that is, if the landing frame is deformed, the score evaluation based on this will result in a large error in the long jump score, meaning that the long jump score is not accurately calculated.

[0072] With the rapid development of deep learning and computer vision technologies, machine vision problems such as image recognition, object detection, semantic segmentation, and pose estimation have found relatively good solutions in the era of convolutional neural networks. Their performance on some public datasets such as ImageNet and MS COCO (Microsoft Common Objects in Context) is already astonishing. Building on this, the increased computing power and decreased cost of parallel computing devices such as GPUs (Graphics Processing Units) and NPUs (Neural-network Processing Units) will lead to the increasingly widespread application of various models in scenarios such as autonomous driving, behavior analysis, and facial recognition.

[0073] In view of this, the present invention provides a long jump evaluation method, aiming to improve the efficiency of long jump evaluation and the teaching efficiency of physical education classes by utilizing computer vision technology and a cyclic state machine. Specifically, the long jump evaluation method provided by the present invention uses deep learning algorithms such as human detection, posture estimation, and foot segmentation to analyze the various states of the test subject, and uses a cyclic state machine designed according to the various stages of the long jump to perform long jump evaluation. Figure 1 This is a flowchart illustrating the long jump evaluation method provided by the present invention, as shown below. Figure 1 As shown, the main implementing body of this method is a long jump assessment system. This system consists of network cameras deployed in specific locations, scale lines painted in the long jump assessment area, and servers. It has low deployment costs and requires no manual maintenance after successful deployment. It can be applied to daily long jump physical education classes, greatly improving the teaching efficiency of physical education classes. The method includes:

[0074] Step 110: Determine the coordinates of the video to be tested and the scale lines within the long jump evaluation area;

[0075] Specifically, before conducting a long jump assessment, it is necessary to first determine the area where the testers will conduct the long jump assessment, i.e., the long jump assessment area. It is also necessary to determine the video of the testers conducting the long jump assessment in this long jump assessment area, as well as the coordinates of the scale lines used for performance evaluation in this long jump assessment area, i.e., the video to be tested and the coordinates of the scale lines under the long jump assessment area.

[0076] Figure 2 This is a schematic diagram of the scale distribution under the long jump evaluation area provided by the invention, such as... Figure 2 As shown, the coordinates of each scale line in the scale distribution can be obtained by calibrating the long jump evaluation area. That is, the scale lines can be calibrated within the long jump evaluation area using a long jump scale calibration algorithm, and the coordinates of the scale lines can be determined. However, considering that in traditional schemes, when calibrating the long jump evaluation area, only the take-off line is usually calibrated, this can easily lead to large errors in subsequent processes such as take-off judgment, personnel tracking, and performance evaluation, resulting in low accuracy of the results, this embodiment of the invention calibrates the take-off area, evaluation area, and interference area in addition to the scale lines and take-off line when calibrating the long jump evaluation area. This achieves a fine division of the long jump evaluation area and overcomes the shortcomings of traditional schemes, such as inaccurate take-off judgment due to simplistic division and low accuracy of performance calculation.

[0077] Furthermore, the region labeling process for the long jump evaluation area can be completed using a region labeling model. Before this, the region labeling model needs to be pre-trained. The training process includes the following steps: First, a large dataset for training is collected. Each image in this dataset needs to be labeled using Labelme (an image annotation tool), that is, each tick mark in the image needs to be marked using a rectangle. Then, the initial region labeling model can be trained based on the labeled dataset to obtain the trained region labeling model. It should be noted that the region labeling model here is built on the basis of an object detection model (such as the YOLOX object detection model).

[0078] During inference, the above process first sorts the output of the initial region calibration model according to the X and Y axes, identifies abrupt changes in distance, and divides the boxes into upper and lower rows. Then, false alarm filtering is performed: for the upper and lower rows, two boxes with an IOU (Intersection over Union) greater than a preset IOU threshold are removed. Next, missing detection completion is performed: for the upper and lower rows, the width of each box and the spacing between them are determined and used as the basis for determining whether any boxes are missing. Then, all boxes are traversed from left to right, and the number of missing boxes between them is determined based on the spacing and width of each box. Then, linear interpolation is performed based on the information of the two missing boxes on the left to complete the missing boxes. Finally, it is checked whether the number of completed boxes meets the scale setting, i.e., whether there are 61 boxes on each side, for a total of 122 boxes.

[0079] Step 120: Determine the landing frame from each frame in the video to be tested;

[0080] Specifically, in step 110, based on the determined video to be tested within the long jump evaluation area, step 120 can be executed to determine the landing frame from each frame of the video to be tested. This process specifically includes the following steps:

[0081] First, human detection can be performed on the first frame of the video to be tested to obtain the human detection result of the first frame. The first frame here can be understood as the current frame or the initial frame of the video to be tested. The human detection of the first frame is actually the human detection of the take-off area under the pre-marked long jump evaluation area to detect whether there is a person in the take-off area.

[0082] Then, based on the human body detection results of the first frame, it can be determined whether the tester is in a ready state. Specifically, if the human body detection results of the first frame indicate that someone is in the take-off area and the number of people is 1, that is, if the human body detection results of the first frame indicate that there is only one person in the take-off area, it can be determined that the tester has entered the preparation stage of the long jump evaluation process, that is, the tester is in a ready state.

[0083] Subsequently, with the tester in a ready state, the movement speed of the tester's feet can be obtained from each frame of the video to be tested. Specifically, the movement speed of the tester's foot bones on the two-dimensional image can be determined from several consecutive frames after the first frame of the video to be tested, and the tester can be determined from this movement speed to determine whether the tester has entered the take-off phase, that is, whether the tester has taken off.

[0084] Subsequently, when the moving speed indicates that the tester has jumped, that is, when the tester is in the jumping state, the moving speed of the tester's feet can be obtained again according to each frame in the video to be tested. That is, the moving speed of the tester's foot bone points can be determined according to the video frames after the jumping frame corresponding to the jumping state in the video to be tested, and the tester can be judged whether the tester is in the landing state based on this moving speed, that is, whether the tester has landed.

[0085] Finally, based on the landing state, the landing frame can be determined from each frame in the video under test. In other words, the landing frame corresponding to the landing state can be determined from the video under test based on the landing state.

[0086] Step 130: Based on the coordinates of the foot bone points of the person in the landing frame, the landing frame is segmented into foot parts to obtain the coordinates of multiple foot contour points in the landing frame.

[0087] Specifically, after step 120, which determines the landing frame from each frame in the video to be tested, step 130 can be executed. Based on the coordinates of the foot bone points of the person in the landing frame, the landing frame is segmented into foot parts to obtain the coordinates of multiple foot contour points in the landing frame. The specific process may include the following steps:

[0088] First, it is necessary to determine the coordinates of the foot bones of the person in the landing frame, that is, the coordinates of the foot bones of the test person in the landing frame. This can be obtained by estimating the posture of the test person based on the landing frame. In other words, the coordinates of the foot bones of the test person can be obtained by estimating the posture of the person in the landing frame.

[0089] Then, using the coordinates of the foot bone points of the person in the landing frame as a reference, the landing frame can be segmented into feet to obtain the coordinates of the foot contour points of the tester in the landing frame. Specifically, firstly, the bounding box of the foot bone points of the tester in the landing frame can be determined based on the coordinates of the foot bone points of the tester in the landing frame. Then, the foot can be segmented based on the bounding box of the foot bone points to obtain multiple foot contour points of the tester in the landing frame, as well as the coordinates of each foot contour point.

[0090] Step 140: Based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines, the long jump performance is evaluated.

[0091] Specifically, after obtaining the coordinates of multiple foot contour points in the landing frame through the above steps, step 140 can be executed. Based on the coordinates of these multiple foot contour points and the coordinates of the scale lines under the long jump evaluation area, the long jump performance is evaluated to determine the tester's long jump performance. The specific process includes the following steps:

[0092] First, based on the coordinates of multiple foot contour points in the landing frame, and by using the long jump scale calibration algorithm to obtain the coordinates of each scale line in the scale distribution with a spacing of 5 cm between 0-300 cm in the long jump evaluation area, the scale interval of each foot contour point in the scale distribution can be determined, and thus the score outside the interval in the long jump result can be obtained.

[0093] Subsequently, the test score can be determined based on the upper or lower limit of the scale interval where each foot contour point is located, and the coordinates of each foot contour point. The score outside the interval is then superimposed on this to obtain the test subject's long jump score. Specifically, the score within the interval can be obtained by subtracting the x or y coordinate of the corresponding foot contour point from the x or y coordinate of the upper limit of the scale interval. Then, the score within the interval is subtracted from the upper limit of the interval (which serves as the score outside the interval). Alternatively, the score within the interval can be obtained by subtracting the x or y coordinate of the lower limit of the scale interval from the x or y coordinate of the foot contour point. Then, the score within the interval is added to the lower limit of the interval (which serves as the score outside the interval).

[0094] In addition, it is worth noting that, considering that the landing frame may be obscured due to the foot area, making it impossible to calculate the long jump score, this embodiment of the invention designs a delayed score evaluation strategy. In this case, the landing frame is stopped from being used for score evaluation, and video frames after the landing frame where the foot area is not obscured are used. This avoids the situation where the long jump score cannot be calculated, ensures the accuracy of the calculated long jump score, and improves the flexibility of the score calculation process.

[0095] The long jump evaluation method provided by this invention determines the test video and scale coordinates within the long jump evaluation area. From each frame of the test video, the landing frame is determined. Based on the coordinates of the foot bone points of the person in the landing frame, the landing frame is segmented to obtain the coordinates of multiple foot contour points in the landing frame. Based on the coordinates of these multiple foot contour points and the scale coordinates, the long jump performance is evaluated. This overcomes the inaccuracy of traditional long jump performance calculations. By analyzing the state of the test personnel through human detection, posture estimation, and foot segmentation, the corresponding landing frame is determined, and the performance is evaluated accordingly, improving the accuracy of long jump performance calculation. Furthermore, the use of computer vision technology and a cyclic state machine for long jump evaluation also improves the efficiency of long jump evaluation and the teaching efficiency of physical education classes.

[0096] Based on the above embodiments, Figure 3 This is a flowchart illustrating step 140 of the long jump evaluation method provided by the present invention, as shown below. Figure 3 As shown, step 140 includes:

[0097] Step 141: Based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines, determine the scale interval where the multiple foot contour points are located.

[0098] Step 142: Determine the person's long jump performance based on the upper or lower limit of the scale interval and the distance between multiple foot contour points and the upper or lower limit of the scale interval.

[0099] Specifically, in step 140, the long jump performance evaluation process, based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines, may include the following steps:

[0100] Step 141: First, based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines, determine the scale interval in the scale distribution under the long jump evaluation area where each foot contour point falls. That is, based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines under the long jump evaluation area, determine the scale interval in which each foot contour point is located. Specifically, if the horizontal or vertical coordinate of the foot contour point is greater than the horizontal or vertical coordinate of a certain scale line in the scale distribution, but less than or equal to the horizontal or vertical coordinate of the adjacent scale line of that scale line, it can be determined that the foot contour point is in the scale interval formed by these two scale lines.

[0101] Step 142: Subsequently, based on the spacing between the scale lines in the scale distribution, the upper or lower limit of the scale interval, and the coordinates of each foot contour point, the distance between each foot contour point and the upper or lower limit of the scale interval can be determined. The spacing between the scale lines here has been determined when the long jump evaluation area is calibrated. In this embodiment of the invention, the spacing is 5 centimeters, which means that the distance between each foot contour point and the upper or lower limit of the scale interval can be determined according to the strategy of proportional division.

[0102] After this, the long jump score of the tester can be determined based on the upper or lower limit of the scale interval where each foot contour point is located, and the distance between each foot contour point and the upper or lower limit of the scale interval. Specifically, this can be done by adding the distance between the corresponding foot contour point and the lower limit of the scale interval to the lower limit of the scale interval where each foot contour point is located, or by subtracting the distance between the corresponding foot contour point and the upper limit of the scale interval where each foot contour point is located from the upper limit of the scale interval. The long jump score corresponding to each foot contour point can then be obtained. The minimum long jump score is selected from the long jump scores corresponding to each foot contour point as the tester's long jump score.

[0103] In this embodiment of the invention, the score within the interval is determined by a strategy of proportional division, and the score outside the interval is superimposed on this basis to obtain the long jump score of the tester. By superimposing the scores inside and outside the interval, the long jump score is evaluated, ensuring the accuracy and precision of the long jump score calculation.

[0104] The following example illustrates the evaluation process for the long jump results:

[0105] Figure 4 This is an example diagram illustrating the evaluation process of long jump performance provided by the present invention, such as... Figure 4 As shown, firstly, based on the coordinates of the foot contour point and the coordinates of the scale line, it can be determined that the foot contour point is located in the scale interval ([x2,x1]) formed by the 2.3-meter equidistant line and the 2.25-meter equidistant line. Then, the distance between the foot contour point and the upper or lower limit of the scale interval can be determined using the strategy of proportional division. After that, the long jump performance of the tester can be determined based on the distance between the foot contour point and the upper or lower limit of the scale interval, as well as the upper or lower limit of the scale interval.

[0106] The formula for calculating the long jump result corresponding to any foot contour point is as follows:

[0107]

[0108] Where L is the long jump result corresponding to the foot contour point, x2 is the abscissa corresponding to the lower limit of the scale interval where the foot contour point is located, x1 is the abscissa corresponding to the upper limit of the scale interval where the foot contour point is located, x is the abscissa of the foot contour point, and 5 represents the spacing between the scale lines in the scale distribution.

[0109] Based on the above embodiments, Figure 5 This is a flowchart illustrating step 141 of the long jump evaluation method provided by the present invention, as shown below. Figure 5 As shown, step 141 includes:

[0110] Step 141-1: Perform curve fitting on multiple foot contour points in the landing frame, and sparsify the foot contour curves obtained from the curve fitting.

[0111] Step 141-2: Based on the coordinates of the multiple foot contour points after sparsification and the coordinates of the scale lines, determine the scale interval where the multiple foot contour points are located.

[0112] Specifically, step 141, which involves determining the scale interval where multiple foot contour points are located based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines, includes the following steps:

[0113] First, after obtaining multiple foot contour points in the landing frame by executing step 141-1, curve fitting can be performed on these multiple foot contour points to fit the multiple foot contour points of the tester in the landing frame into a curve, thereby obtaining the foot contour curve of the tester in the landing frame. Specifically, polygonal curve fitting can be performed on multiple foot contour points in the landing frame to obtain the foot contour curve of the tester in the landing frame. Then, this foot contour curve can be sparsified to obtain accurate multiple foot contour points of the tester in the landing frame.

[0114] Then, by executing step 141-2, the scale interval where these multiple foot contour points are located can be determined based on the precise coordinates of the multiple foot contour points obtained in the previous step and the coordinates of the scale lines under the long jump evaluation area. The specific process has been explained in detail above and will not be repeated here.

[0115] Based on the above embodiments, the coordinates of the foot bone points are determined according to the following steps:

[0116] Based on the attitude estimation model, the attitude of the person in the landing frame is estimated to obtain the coordinates of the foot bone points of the person in the landing frame.

[0117] The posture estimation model is trained based on the first sample image, the region label of the long jump evaluation area in the first sample image and the human skeleton point label, with the constraint that the person is in the long jump evaluation area.

[0118] Specifically, in step 130, the coordinates of the foot bones of the person in the landing frame, that is, the coordinates of the foot bones of the test person in the landing frame, can be determined based on the following steps:

[0119] First, the posture of the person in the landing frame can be estimated to obtain the coordinates of the foot bones of the person in the landing frame. This process can be achieved with the help of a posture estimation model. That is, the landing frame can be input into the posture estimation model, and the posture estimation model can estimate the posture of the person in the input landing frame. Finally, the coordinates of the foot bones of the person in the landing frame are output by the posture estimation model, that is, the coordinates of the foot bones of the person in the landing frame.

[0120] Before inputting the landing frame into the attitude estimation model, the attitude estimation model can be pre-trained using the first sample image, with the constraint that the person is in the long jump evaluation area of ​​the corresponding scene for long jump evaluation.

[0121] The training process of the pose estimation model includes: First, collecting a large number of first sample images to form a training dataset, and labeling the long jump evaluation region and human skeleton points in each first sample image in the training dataset to form region labels for the long jump evaluation region and human skeleton point labels. It should be noted that the first sample images here must be images of people performing long jump evaluations in the long jump evaluation region of the corresponding scene, which can be collected from the MS COCO public dataset; then, the initial pose estimation model can be trained based on the first sample images, as well as the region labels and human skeleton point labels of the long jump evaluation region in the first sample images, to obtain the trained pose estimation model.

[0122] It is worth noting that the initial pose estimation model here is built on the SimDR (Simple Disentagled coordinate Representation) single-person pose estimation framework based on the one-dimensional heatmap paradigm.

[0123] Furthermore, in this embodiment of the invention, the MS COCO public dataset (17 points) and the human skeleton points (30 points) of the person performing the long jump evaluation in the corresponding scenario's long jump evaluation area are used to train the initial posture estimation model, which can produce a posture estimation model that can output 30 points.

[0124] Based on the above embodiments, Figure 6 This is a flowchart illustrating step 130 of the long jump evaluation method provided by the present invention, as shown below. Figure 6 As shown, step 130 includes:

[0125] Step 131: Determine the bounding box of the foot bone points based on their coordinates;

[0126] Step 132: Based on the foot segmentation model, the bounding box of the foot bone points is segmented to obtain the coordinates of multiple foot contour points in the landing frame.

[0127] The foot segmentation model is trained based on the bounding boxes of foot bone points in the second sample image, as well as the region labels and foot contour point labels of the long jump evaluation area in the second sample image, with the constraint that the person is in the long jump evaluation area.

[0128] Specifically, step 130 involves segmenting the landing frame into foot sections based on the coordinates of the foot bone points of the person in the landing frame, thereby obtaining the coordinates of multiple foot contour points of the test person in the landing frame. This process includes the following steps:

[0129] First, in step 131, the bounding box of the foot bone points of the person in the landing frame can be determined based on the coordinates of the foot bone points of the person in the landing frame, i.e., the coordinates of the foot bone points of the test person in the landing frame.

[0130] Then, step 132 is executed. The foot segmentation model can be used to segment the bounding box of the foot bone points to obtain the coordinates of multiple foot contour points in the landing frame. Specifically, the bounding box of the foot bone points is first input into the foot segmentation model, and then the foot segmentation model segments the bounding box of the input foot bone points. Finally, the coordinates of multiple foot contour points of the tester in the landing frame can be obtained from the foot segmentation model.

[0131] It should be noted that before inputting the bounding boxes of the foot bone points into the foot segmentation model, the foot segmentation model can be pre-trained using the bounding boxes of the foot bone points in the second sample image, with the constraint that the person is in the long jump evaluation area of ​​the corresponding scene for long jump evaluation.

[0132] The training process of the foot segmentation model includes: First, collecting a large number of second sample images to build a training dataset, and labeling the long jump evaluation region and foot contour points in each second sample image in the training dataset to form region labels for the long jump evaluation region and foot contour point labels. It should be noted that the second sample images here must be images of people performing long jump evaluations in the long jump evaluation region of the corresponding scene, which can be collected from the LIP (Look into Person) human parsing dataset; then, the initial foot segmentation model can be trained based on the bounding boxes of the foot bone points in the second sample images, as well as the region labels and foot contour point labels of the long jump evaluation region in the second sample images, so as to obtain the trained foot segmentation model.

[0133] It is worth noting that the initial foot segmentation model here is built on the semantic segmentation framework based on HRNet (High-Resolution Net).

[0134] It should be noted that in this embodiment of the invention, the LIP human body parsing dataset and the foot contour points of the personnel performing long jump evaluation in the corresponding long jump evaluation area are used to train the initial foot segmentation model. The bounding box obtained by expanding the bounding box of the foot bone points in the second sample image by a certain proportion is used as the input of the initial foot segmentation model to train a model that can effectively segment the foot region.

[0135] In addition to obtaining the coordinates of foot contour points using the foot segmentation model mentioned above, this embodiment of the invention can also obtain the coordinates of foot contour points in other ways based on the foot bone points. For example, specific bone points can be selected from the foot bone points of the tester, that is, bone points that can represent the foot contour of the tester can be selected from the foot bone points of the tester in the landing frame. These bone points are the foot contour points of the tester, and their coordinates are the coordinates of multiple foot contour points of the tester in the landing frame.

[0136] Based on the above embodiments, Figure 7 This is a flowchart illustrating step 120 of the long jump evaluation method provided by the present invention, as follows: Figure 7 As shown, step 120 includes:

[0137] Step 121: Based on the human detection results of the first frame in the video to be tested, determine the state of the person, and when the state is in the ready state, obtain the first movement speed of the person's foot bone points based on each frame in the video to be tested.

[0138] Step 122: Given that the first moving speed represents the person being in a jumping state, the second moving speed of the person's foot bone points is obtained based on each frame in the video to be tested.

[0139] Step 123: If the second moving speed is less than the preset landing speed threshold, determine that the person is in a landing state, and determine the landing frame corresponding to the landing state from each frame in the video to be tested.

[0140] Specifically, step 120, the process of determining the landing frame from each frame in the video to be tested, includes the following steps:

[0141] In this embodiment of the invention, a cyclic state machine is used for state transitions to determine the landing state, and the landing frame corresponding to the landing state is determined from the video under test. Figure 8 This is a schematic diagram of the cyclic state machine in the long jump evaluation method provided by the present invention, as shown below. Figure 8 As shown, the long jump evaluation system is in a waiting state at the beginning. Then, according to the preset detection interval, the human detection model is called to perform human detection, that is, to execute step 121. The human detection model can be used to perform human detection on the first frame of the video to be tested, so as to detect whether there is a person in the take-off area under the long jump evaluation area, thereby obtaining the human detection result of the first frame. Specifically, the first frame can be input into the human detection model, and the human detection model performs human detection on the input first frame, and finally obtains the human detection result of the first frame output by the human detection model.

[0142] Before inputting the first frame into the human detection model, the human detection model can be pre-trained using a third sample image, with the constraint that the person is in the long jump evaluation area of ​​the corresponding scene.

[0143] The training process of the human detection model includes: First, collecting a large number of third sample images to form a training dataset, and labeling the long jump evaluation region and human body region in each third sample image in the training dataset to form region labels for the long jump evaluation region and human body bounding box labels. The third sample images here can be the same as or different from the first sample images mentioned above; this embodiment of the invention does not specifically limit this, but the third sample images must be images of a person in the long jump evaluation region of the corresponding scene, which can be collected from the MS COCO public dataset. Then, based on the third sample images and the region labels and human body bounding box labels of the long jump evaluation region in the third sample images, the initial human detection model can be trained to obtain the trained human detection model. It is worth noting that the initial foot segmentation model here is built on the YOLOX object detection framework.

[0144] Furthermore, after obtaining the human detection result of the first frame, it is possible to determine whether the tester is in a ready state based on this result. Specifically, if the human detection result of the first frame indicates that someone is in the take-off area under the long jump evaluation area, and the number of people in the take-off area is 1, that is, if there is exactly one person in the take-off area under the long jump evaluation area, it can be preliminarily determined that the tester has entered the ready state. At this time, in order to further confirm the state of the tester, human detection can continue. That is, based on the first frame, human detection can be performed again at a preset interval. That is, according to the preset detection interval, the human detection model is called to perform human detection to obtain human detection results. If the human detection results of a preset number of consecutive times indicate that there is exactly one person in the take-off area under the long jump evaluation area, it is determined that the tester is in a ready state.

[0145] It should be noted that the preset detection interval here can be set according to the actual situation, for example, it can be 20 frames, 25 frames, 30 frames, etc., and as a preferred embodiment of the present invention, the preset detection interval is selected as 25 frames, and correspondingly, the preset frame is also 25 frames; the preset number of times can also be set according to the actual situation, it can be 2 times, 3 times, 4 times, etc., and as a preferred embodiment of the present invention, the preset number of times is determined to be 3 times.

[0146] Once it is confirmed that the tester is in a ready state, the movement speed of the tester's foot bone points, i.e. the first movement speed, can be obtained based on each frame in the video to be tested. Specifically, this process can be based on several consecutive frames after the first frame in the video to be tested to determine the first movement speed of the tester's foot bone points on the two-dimensional image.

[0147] Then, step 122 is executed. When the first moving speed indicates that the tester has entered the jumping phase, that is, when the tester is in the jumping state, the moving speed of the tester's foot bone points can be continuously obtained according to each frame in the video to be tested, that is, the second moving speed. Specifically, the second moving speed of the tester's foot bone points can be determined according to the video frames after the jumping frame corresponding to the jumping state in the video to be tested.

[0148] After that, by executing step 123, the landing status of the tester can be determined based on the second moving speed. That is, if the second moving speed is less than the preset landing speed threshold, it is determined that the tester is in the landing status. Based on the landing status, the landing frame can be determined from each frame of the video under test. In other words, the landing frame corresponding to the landing status can be determined from the video under test based on the landing status.

[0149] Correspondingly, if the second moving speed is greater than or equal to the preset landing speed threshold, it can be determined that the test personnel have not yet landed, that is, they are still in the air. The second moving speed needs to be continuously acquired until it is less than the preset speed threshold, then it can be determined that the test personnel have landed.

[0150] In this embodiment of the invention, the second moving speed is determined by the video frame after the jump frame, and the tester is judged to be in a landing state based on the second moving speed and a preset speed threshold. This avoids the shortcomings of traditional solutions, which are difficult to capture and measure due to the difficulty in judging the landing state (angles formed by the hips, knees, ankles, etc.), resulting in large judgment errors and low accuracy. By judging the landing state through a normalized preset landing speed threshold, the judgment process is simplified while improving the judgment efficiency and accuracy.

[0151] Furthermore, the preset landing speed threshold in this embodiment of the invention is bound to the scale distribution of the long jump evaluation area. In other words, it can be determined based on the coordinates of the scale lines under the long jump evaluation area. This allows the landing state judgment threshold to be independent of the actual scenario, that is, it can be independent of the deployment scenario of the long jump evaluation system, thus improving the robustness of the entire system.

[0152] Based on the above embodiments, Figure 9 This is a flowchart illustrating step 122 of the long jump evaluation method provided by the present invention, as shown below. Figure 9 As shown, step 122 includes:

[0153] Step 122-1: Determine the jump frame corresponding to the jump state from each frame in the video to be tested;

[0154] Step 122-2: Based on the coordinates of the foot bone points of the person in the take-off frame and the coordinates of the foot bone points of the person in the second frame after the take-off frame in the video to be tested, determine the displacement of the foot bone points of the person.

[0155] Step 122-3: Based on the coordinates of the person's foot bones in the second frame, the displacement of the person's foot bones, and the coordinates of the take-off line in the long jump evaluation area, the take-off state is verified. If the verification is successful, the second movement speed of the person's foot bones is obtained based on each frame after the second frame in the video to be tested.

[0156] Considering that traditional long jump evaluations are mostly designed with a serial evaluation mechanism, under this evaluation mechanism, the take-off, which has a significant impact on the entire evaluation process, is very likely to cause the entire evaluation process to end prematurely if an abnormal trigger occurs. For example, if the abnormal trigger is caused by skeletal shaking, the entire evaluation process will end directly.

[0157] In view of this, an embodiment of the present invention designs a cyclic state machine, that is, after initially determining that the person is in the take-off state, a secondary confirmation mechanism is introduced. The mechanism determines whether the person is truly in the take-off state, i.e. whether the person has actually taken off, by measuring the displacement of the foot bone points of the tester and the positional relationship between the foot bone points and the take-off line.

[0158] Specifically, step 122, the process of obtaining the second movement velocity of the person's foot bone points based on each frame in the video to be tested, may include the following steps:

[0159] Step 122-1: When the first transfer speed indicates that the test personnel have entered the take-off phase, that is, when it is initially determined that the test personnel are in the take-off state, the take-off frame corresponding to the take-off state can be determined from each frame in the video to be tested.

[0160] Step 122-2: The coordinates of the tester's foot bones in the jump frame can be used as a reference. Combined with the coordinates of the tester's foot bones in the second frame after the jump frame in the video to be tested, the displacement of the tester's foot bones can be determined. That is, the distance the tester's foot bones move in a certain direction can be calculated by using the coordinates of the tester's foot bones in the two frames. Here, the second frame can be understood as the video frame after a certain time window from the first frame.

[0161] Step 122-3: First, it is necessary to determine the coordinates of the foot bones of the tester in the second frame. This can be obtained by calling the attitude estimation model to estimate the attitude of the tester in the second frame. Then, the coordinates of the foot bones of the tester in the second frame, the displacement of the foot bones of the tester determined in step 122-2, and the coordinates of the take-off line under the pre-marked long jump evaluation area can be used to verify the take-off state of the tester initially determined, so as to verify whether the tester has actually taken off.

[0162] Furthermore, if the verification passes, i.e., if the tester does indeed jump, it can be determined that the tester is in the air. At this point, the second movement speed of the tester's foot bones can be continuously obtained from each frame after the second frame in the video to be tested.

[0163] Correspondingly, if the verification fails, that is, if it is verified that the initial jump state was caused by an abnormal trigger, the jump at this time is not a real jump, so the tester's state needs to be reset to the ready state.

[0164] In this embodiment of the invention, the take-off state is verified by the displacement of the foot bone points of the tester and the positional relationship between the foot bone points and the take-off line. This solves the problem in the traditional solution where abnormal triggering caused by the shaking of the human bone points leads to the direct termination of the evaluation process, and ensures the meticulousness of the state transition link based on the cyclic state machine.

[0165] Based on the above embodiments, in step 122-3, the take-off state is verified based on the coordinates of the person's foot bones in the second frame, the displacement of the person's foot bones, and the coordinates of the take-off line under the long jump evaluation area, including:

[0166] If the displacement of the person's foot bones is greater than the preset displacement threshold, and the coordinates of the person's foot bones and the take-off line coordinates in the second frame reflect that the person's foot bones have crossed the take-off line of the long jump evaluation area, then the verification of the take-off state is determined to be passed.

[0167] If the displacement of the athlete's foot bones is less than or equal to the preset displacement threshold, or if the coordinates of the athlete's foot bones and the take-off line coordinates in the second frame indicate that the athlete's foot bones have not crossed the take-off line of the long jump evaluation area, the check for the take-off state is determined to have failed, and the take-off state is reset to the ready state.

[0168] Specifically, in step 122-3, the process of verifying the take-off status based on the coordinates of the person's foot bones in the second frame, the displacement of the person's foot bones, and the coordinates of the take-off line under the long jump evaluation area includes the following two cases:

[0169] Firstly, if the displacement of the tester's foot bones exceeds the preset displacement threshold, and the coordinates of the tester's foot bones and the take-off line in the second frame indicate that the tester's foot bones have crossed the take-off line of the long jump evaluation area, that is, if the displacement of the tester's foot bones exceeds the preset displacement threshold, and the second frame indicates that the tester's foot bones have crossed the take-off line, then it can be determined that the tester's take-off state has passed the verification, meaning that the tester has actually taken off and is in the air.

[0170] Secondly, if the displacement of the personnel's foot bones is less than or equal to the preset displacement threshold, or if the coordinates of the personnel's foot bones and the take-off line coordinates in the second frame indicate that the personnel's foot bones have not crossed the take-off line of the long jump evaluation area, that is, if the personnel's foot bones are less than or equal to the preset displacement threshold, or if the second frame indicates that the personnel's foot bones have not crossed the take-off line, it can be determined that the verification of the personnel's take-off state has failed, that is, the personnel have not taken off. In this case, the personnel's state needs to be reset to the ready state.

[0171] Based on the above embodiments, Figure 10 This is a schematic diagram of the process for determining the take-off state provided by the present invention, as shown below. Figure 10 As shown, the takeoff state is determined based on the following steps:

[0172] Step 1010: Based on the first moving speed, determine the horizontal moving speed component and the vertical moving speed component of the person's foot bone points.

[0173] Step 1020: If the horizontal and vertical moving velocity components are both greater than the preset soft velocity threshold, and the horizontal or vertical moving velocity component is greater than the preset hard velocity threshold, then determine that the person is in the jumping state.

[0174] The preset soft speed threshold is less than the preset hard speed threshold.

[0175] Specifically, the process of determining whether the test subject is in a jumping state based on the first moving speed can include the following steps:

[0176] Step 1010: First, based on the first moving speed, the horizontal moving speed component and the vertical moving speed component of the tester's foot bone points can be determined. That is, the first moving speed can be decomposed into horizontal moving speed component and vertical moving speed component, thereby obtaining the horizontal moving speed component and the vertical moving speed component of the tester's foot bone points.

[0177] Step 1020: Based on the horizontal and vertical movement component velocities, the take-off state can be determined. Specifically, if the horizontal and vertical movement component velocities of the tester's foot bones are both greater than a preset soft velocity threshold, and either the horizontal or vertical movement component velocities are greater than a preset hard velocity threshold, it can be preliminarily determined that the tester is in the take-off state.

[0178] It is worth noting that the preset soft speed threshold and the preset hard speed threshold can be set according to actual needs, but the preset soft speed threshold must be less than the preset hard speed threshold.

[0179] Correspondingly, if the horizontal and / or vertical velocity components of the tester's foot bones are less than or equal to a preset soft velocity threshold, or if both the horizontal and vertical velocity components are less than or equal to a preset hard velocity threshold, it can be determined that the tester has not yet entered the take-off phase and is still in a ready state.

[0180] In this embodiment of the invention, by decomposing the first moving velocity and utilizing its moving velocity components in two directions, along with preset soft velocity thresholds and preset hard velocity thresholds, the take-off state is determined. This avoids the shortcomings of traditional solutions, which suffer from difficulties in judging the take-off state due to the difficulty in capturing and measuring the angles formed by the hips, knees, ankles, etc., leading to large judgment errors and low accuracy. By using soft and hard velocity thresholds for take-off state judgment, the judgment process is simplified while improving judgment efficiency and accuracy. Furthermore, it also solves the problem in traditional solutions where the overly simplistic take-off rules prevent the capture of the take-off state in certain special take-off postures, greatly improving the efficiency of take-off state judgment and advancing the long jump evaluation process.

[0181] Furthermore, it should be noted that the preset soft speed threshold and preset hard speed threshold in the embodiments of the present invention are both bound to the scale distribution of the long jump evaluation area. In other words, they can be determined according to the coordinates of the scale lines under the long jump evaluation area. This allows the judgment threshold of the take-off state to be independent of the actual scenario, that is, it can be independent of the deployment scenario of the long jump evaluation system, thus improving the robustness of the entire system.

[0182] Based on the above embodiments, when the tester is in a ready state, the long jump tracking process can be started to track the tester. Specifically, after obtaining the pose estimation result of the current frame (the coordinates of the foot bone points of the person in the current frame), the human detection box of the next frame in the video under test will be replaced by the bounding box of the foot bone points determined by the pose estimation result of the current frame and expanded by a certain proportion. In other words, after calling the pose estimation model to estimate the pose of the person in the current frame to obtain the coordinates of the foot bone points of the person in the current frame, and determining the bounding box of the foot bone points of the person in the current frame based on the coordinates of the foot bone points, the bounding box can be expanded by a certain proportion as the human detection box of the next frame.

[0183] In this embodiment of the invention, a top-down tracking strategy is adopted for personnel tracking. This overcomes the shortcomings of traditional solutions that use bottom-up multi-person posture estimation results for personnel tracking, which requires processing the entire image frame for each frame and estimating the posture of all people in the image. This process is not only cumbersome but also likely to result in inaccurate estimation of foot bone points, thus affecting the state transitions in the long jump evaluation process. In this embodiment of the invention, a top-down tracking strategy is adopted for tracking test personnel. Based on a human detection algorithm, personnel in the take-off area below the long jump evaluation area are selected, and personnel are tracked frame by frame based on a posture estimation algorithm. This ensures the accuracy of test personnel tracking. Furthermore, this tracking strategy only needs to process local areas in the video frame, simplifying the system complexity. At the same time, it can more accurately depict the foot bone points of the test personnel, improving the system's operating efficiency.

[0184] Based on the above embodiments, Figure 11 This is a schematic diagram of the long jump violation assessment process provided by the present invention, as shown below. Figure 11 As shown, the method also includes:

[0185] Step 1110: In the ready state, perform pose estimation on the first frame to obtain the coordinates of the foot bone points of the person in the first frame.

[0186] Step 1120: Based on the coordinates of the foot bone points of the person in the first frame, the first frame is segmented into foot parts to obtain the coordinates of multiple foot contour points in the first frame;

[0187] Step 1130: Based on the coordinates of multiple foot contour points in the first frame and the coordinates of the take-off line under the long jump evaluation area, conduct a long jump line violation assessment.

[0188] Specifically, the method provided in this embodiment of the invention, in addition to evaluating long jump performance, can also evaluate the behavior of the tester during the long jump evaluation process. See [link to relevant documentation]. Figure 8 It is known that the evaluation can be conducted by assessing whether the test personnel violated regulations such as stepping on the line, interrupting the test, standing incorrectly, or jumping incorrectly, as well as assessing whether there was interference from personnel. In this embodiment of the invention, the violation of stepping on the line during the long jump evaluation process is evaluated, and the specific process includes the following steps:

[0189] Step 1110: When the state is in the ready state, the first frame can be used for attitude estimation to obtain the coordinates of the foot bone points of the person in the first frame. Specifically, when it is determined that the test person is in the jumping state, the attitude estimation model can be called to perform attitude estimation on the person in the first frame of the test video to obtain the coordinates of the foot contour points of the test person in the first frame. The attitude estimation process has been explained in detail above and will not be repeated here.

[0190] Step 1120: Using the coordinates of the tester's foot bone points in the first frame as a reference, the first frame can be segmented to obtain the coordinates of multiple foot contour points of the tester in the first frame. Specifically, firstly, the bounding box of the tester's foot bone points in the first frame can be determined based on the coordinates of the tester's foot bone points in the first frame. Then, the foot segmentation model can be called to segment the bounding box of the tester's foot bone points in the first frame, thereby obtaining multiple foot contour points of the tester in the first frame, as well as the coordinates of each foot contour point. The process of foot segmentation has been explained in detail above and will not be repeated here.

[0191] Step 1130 involves determining whether the tester violated the rule by stepping on the line, based on the coordinates of multiple foot contour points of the tester in the first frame and the coordinates of the take-off line under the long jump evaluation area. Specifically, this can be done by using the coordinates of multiple foot contour points of the tester in the first frame and the coordinates of the take-off line under the long jump evaluation area to calculate the distance between each foot contour point and the take-off line under the long jump evaluation area. It is worth noting that the distance here is a signed distance value. A negative distance value indicates that the corresponding foot contour point is behind (left) and the take-off line is in front (right), meaning that the distance between the corresponding foot contour point and the take-off line under the long jump evaluation area indicates that the tester did not step on the line. Only when the distance between all foot contour points and the take-off line under the long jump evaluation area indicates that the tester did not step on the line, or when the distance between the foot contour point closest to the take-off line and the take-off line indicates that the tester did not step on the line, can it be determined that the tester did not violate the rule by stepping on the line.

[0192] Correspondingly, when the distance value is positive and greater than the preset violation distance threshold, that is, when the corresponding foot contour point is in front (right) and the take-off line is behind (left), in other words, if the distance between the corresponding foot contour point and the take-off line below the long jump evaluation area indicates that the tester has stepped on the line, it can be determined that the tester has violated the rule by stepping on the line. Conversely, if the distance between any foot contour point and the take-off line below the long jump evaluation area indicates that the tester has stepped on the line, it can be determined that the tester has not violated the rule by stepping on the line. The preset violation distance threshold can be set according to the actual situation, for example, it could be 0.5 cm, 1 cm, 1.5 cm, etc.

[0193] Based on the above embodiments, Figure 12 This is a schematic diagram of the long jump take-off violation assessment process provided by the present invention, as shown below. Figure 12 As shown, the method also includes:

[0194] Step 1210: Based on the jump frame in the video to be tested and the third frame before the jump frame, obtain the third movement speed of the foot bone points of the person's left and right feet.

[0195] Step 1220: Based on the third movement speed of the foot bones of the left and right feet of the person and the preset violation speed threshold, the long jump take-off violation assessment is carried out.

[0196] Specifically, the process of assessing take-off violations during the long jump evaluation may include the following steps:

[0197] Step 1210: If the tester is in the jumping state, the starting frame is determined from the frames in the video to be tested based on the jumping state. The starting frame is used as a timestamp to backtrack the video frames in the video to be tested at certain time intervals. Based on the backtracked video frames and the starting frame, the movement speed of the foot bones of the tester's left and right feet is determined. Specifically, the starting frame in the video to be tested is used as a reference, and the third frame before the starting frame is used to determine the movement speed of the foot bones of the tester's left and right feet when jumping, i.e., the third movement speed. The third frame here can be understood as the video frame backtracked in the video to be tested at certain time intervals with the starting frame as a timestamp.

[0198] Step 1220: Based on the third movement speed and the preset violation speed threshold, it can be determined whether the tester violated the jump rule, that is, whether the tester jumped on one foot. Specifically, if the difference between the third movement speed of the left foot bone point and the third movement speed of the right foot bone point when the tester jumps is greater than or equal to the preset violation speed threshold, that is, if the difference in the jump speed of the two feet is too large, it can be determined that the tester jumped on one foot, and the tester can be judged to have violated the jump rule. The preset violation speed threshold can be set according to the actual situation.

[0199] Correspondingly, if the difference between the third movement speed of the left foot bone point and the third movement speed of the right foot bone point when the tester jumps is less than the preset violation speed threshold, that is, if the difference in the jump speed of the two feet is within the tolerable range, it can be determined that the tester's jump is normal, that is, the tester did not jump on one foot, and it can be determined that the tester did not violate the jump rule.

[0200] It should be noted that, in the process of backtracking the video frames in the video to be tested using the start frame as the timestamp, in addition to judging the start frame violation based on the backtracked video frames, the line violation can also be judged based on this. The specific judgment process has been explained above and will not be repeated here.

[0201] Based on the above embodiments, the method further includes:

[0202] Human detection is performed on the fourth frame after the first frame to obtain the human detection results for the fourth frame.

[0203] If the human detection results in the fourth frame indicate that there is no one in the take-off area under the long jump evaluation area, it is determined that the person's long jump was interrupted and violated the rules.

[0204] Specifically, the process of assessing violations of interruption during the long jump evaluation may include the following steps:

[0205] First, the fourth frame can be determined from the video to be tested by taking the first frame as a reference and at a preset interval of frames. The preset frame can be set according to the actual situation. In this embodiment of the invention, the preset frame is set to 25 frames.

[0206] Subsequently, the human detection model can be invoked to perform human detection on the fourth frame to detect whether there is anyone in the take-off area under the long jump evaluation area in the fourth frame, thereby obtaining the human detection results of the fourth frame.

[0207] Furthermore, if the human detection results in the fourth frame indicate that there is no one in the take-off area under the long jump evaluation area, that is, if the human detection in the fourth frame indicates that there is no one in the take-off area under the long jump evaluation area, it can be determined that the tester has left, i.e., the long jump is interrupted and violated the rules.

[0208] Correspondingly, if the human detection results in the fourth frame indicate that there are people in the take-off area below the long jump evaluation area, it can be determined that the tester did not violate the rules by interrupting the test.

[0209] In addition to the aforementioned violations of stepping on the line, taking off, and interruption, the standing direction of the tester can also be used to determine whether the tester is standing backwards. This can be done by referring to the bone points on the tester's feet. If the bone points on the tester's feet indicate that the tester's standing direction is opposite to the direction of the scale lines under the long jump evaluation area from small to large, it can be determined that the tester is standing backwards. In other words, if the bone points on the toes are in the back (left) and the bone points on the heels are in the front (right), it can be determined that the tester is standing in violation of the rules.

[0210] Correspondingly, if the tester's foot bone points reflect the same direction as the scale lines under the long jump evaluation area from small to large, it can be determined that the tester is not standing in the opposite direction. In other words, if the bone point at the toe is in front (right) and the bone point at the heel is behind (left), it can be determined that the tester is not standing in violation of the rules.

[0211] Based on the above embodiments, at least one of the preset landing speed threshold, preset soft speed threshold, and preset hard speed threshold is determined based on the coordinates of the scale line under the long jump evaluation area.

[0212] Specifically, in the long jump evaluation method provided by the embodiments of the present invention, at least one of the preset landing speed threshold for determining the landing state, the preset soft speed threshold for determining the take-off state, and the preset hard speed threshold is bound to the scale distribution of the long jump evaluation area. In other words, any one or two of these three can be determined based on the coordinates of the scale lines under the long jump evaluation area. Of course, all three can also be determined based on the coordinates of the scale lines under the long jump evaluation area.

[0213] In this embodiment of the invention, binding at least one of the preset landing speed threshold, preset soft speed threshold, and preset hard speed threshold to the coordinates of the scale line under the long jump evaluation area enables the judgment thresholds of the take-off state and landing state to be independent of the actual scenario. That is, it can be independent of the deployment scenario of the long jump evaluation system. This overcomes the shortcomings of traditional solutions where the threshold setting needs to be combined with the actual scenario and there is a lack of unified threshold setting rules, resulting in different thresholds for different scenarios and a very cumbersome threshold setting. While simplifying the threshold setting, it improves the robustness of the entire system.

[0214] Based on the above embodiments, Figure 13 This is a general framework diagram of the long jump evaluation method provided by the present invention, as shown below. Figure 13 As shown, the long jump evaluation system is in a waiting state at the beginning. At this time, the system will announce "Please get ready" and enter the ready state. Then the system will announce "Start" and then proceed to the next stage of the long jump evaluation process.

[0215] First, it is necessary to determine the video to be tested and the coordinates of the scale lines within the long jump evaluation area;

[0216] Subsequently, the landing frame can be determined from each frame of the video under test. Specifically, based on the human detection results of the first frame of the video under test, the state of the person can be determined. If the state is in the ready state, the first moving speed of the person's foot bone points can be obtained based on each frame of the video under test. If the first moving speed indicates that the person is in the take-off state, the second moving speed of the person's foot bone points can be obtained based on each frame of the video under test. If the second moving speed is less than a preset landing speed threshold, it can be determined that the person is in the landing state, and the landing frame corresponding to the landing state can be determined from each frame of the video under test.

[0217] The take-off state here can be determined based on the following steps: based on the first moving speed, determine the horizontal and vertical moving speed components of the person's foot bones; if both the horizontal and vertical moving speed components are greater than a preset soft speed threshold, and either the horizontal or vertical moving speed component is greater than a preset hard speed threshold, then the person is determined to be in the take-off state; the preset soft speed threshold here is less than the preset hard speed threshold.

[0218] It should be noted that at least one of the preset landing speed threshold, preset soft speed threshold, and preset hard speed threshold is determined based on the coordinates of the scale line under the long jump evaluation area.

[0219] Specifically, the process of obtaining the second movement speed of the person's foot bones based on each frame of the test video when the first movement speed indicates that the person is in the take-off state can be as follows: First, when the first movement speed indicates that the person is in the take-off state, determine the take-off frame corresponding to the take-off state from each frame of the test video; second, based on the coordinates of the person's foot bones in the take-off frame and the coordinates of the person's foot bones in the second frame after the take-off frame in the test video, determine the displacement of the person's foot bones; third, based on the coordinates of the person's foot bones in the second frame, the displacement of the person's foot bones, and the coordinates of the take-off line under the long jump evaluation area, verify the take-off state; and if the verification passes, obtain the second movement speed of the person's foot bones based on each frame after the second frame in the test video.

[0220] Furthermore, the process of verifying the take-off state based on the coordinates of the athlete's foot bones in the second frame, the displacement of the athlete's foot bones, and the coordinates of the take-off line in the long jump evaluation area includes two cases: First, if the displacement of the athlete's foot bones is greater than a preset displacement threshold, and the coordinates of the athlete's foot bones in the second frame and the take-off line coordinates reflect that the athlete's foot bones have crossed the take-off line of the long jump evaluation area, then the verification of the take-off state is determined to be successful; Second, if the displacement of the athlete's foot bones is less than or equal to the preset displacement threshold, or if the coordinates of the athlete's foot bones in the second frame and the take-off line coordinates reflect that the athlete's foot bones have not crossed the take-off line of the long jump evaluation area, then the verification of the take-off state is determined to be unsuccessful, and the take-off state is reset to the ready state.

[0221] Subsequently, based on the coordinates of the foot bone points of the person in the landing frame, the landing frame can be segmented into foot parts to obtain the coordinates of multiple foot contour points in the landing frame.

[0222] Here, the coordinates of the foot bones of the person in the landing frame are determined based on the following steps: Based on the attitude estimation model, the attitude of the person in the landing frame is estimated to obtain the coordinates of the foot bones of the person in the landing frame; the attitude estimation model is trained based on the first sample image, the region label of the long jump evaluation area and the human bone point label in the first sample image, with the constraint that the person is in the long jump evaluation area for long jump evaluation.

[0223] The process of segmenting the landing frame based on the coordinates of the foot skeleton points of the person in the landing frame to obtain the coordinates of multiple foot contour points in the landing frame can specifically include: determining the bounding box of the foot skeleton points based on the coordinates of the foot skeleton points; segmenting the bounding box of the foot skeleton points based on the foot segmentation model to obtain the coordinates of multiple foot contour points in the landing frame; the foot segmentation model is trained based on the bounding box of the foot skeleton points in the second sample image, as well as the region label and foot contour point label of the long jump evaluation area in the second sample image, with the constraint that the person is in the long jump evaluation area for long jump evaluation.

[0224] After that, the long jump performance can be evaluated based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines. The specific process may include: determining the scale interval where multiple foot contour points are located based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines; determining the person's long jump performance based on the upper or lower limit of the scale interval and the distance between multiple foot contour points and the upper or lower limit of the scale interval.

[0225] Specifically, the process of determining the scale interval of multiple foot contour points based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines can be as follows: performing curve fitting on multiple foot contour points in the landing frame, and sparsifying the foot contour curve obtained by curve fitting; and determining the scale interval of multiple foot contour points based on the coordinates of the multiple foot contour points after sparsification and the coordinates of the scale lines.

[0226] It should be noted that, in addition to determining the long jump score, the landing status must also be evaluated for violations (evaluating whether the landing exceeded the limits, whether the landing exceeded the time limit, and whether the landing involved a turn). Evaluating whether the landing exceeded the limits means determining whether multiple foot contour points in the landing frame are within the pre-marked limits for score calculation. Evaluating whether the landing involved a turn means determining whether the foot contour points were turned during landing, causing them to face the starting point. Evaluating whether the landing exceeded the time limit means determining whether the landing occurred within the specified time.

[0227] Furthermore, if at least one of the above three conditions occurs, namely, landing outside the bounds, landing with a turn, or landing exceeding the time limit, the system will directly determine that the long jump result is invalid and end the long jump evaluation process after the system announces "Result invalid".

[0228] In addition, the long jump evaluation method provided in this embodiment of the invention can also evaluate violations of behavior during the long jump. The evaluation process for violations of stepping on the take-off line in the long jump can include: when the state is in the ready state, performing posture estimation on the first frame to obtain the coordinates of the foot bone points of the person in the first frame; based on the coordinates of the foot bone points of the person in the first frame, segmenting the first frame into foot parts to obtain the coordinates of multiple foot contour points in the first frame; and evaluating violations of stepping on the take-off line in the long jump based on the coordinates of the multiple foot contour points in the first frame and the coordinates of the take-off line under the long jump evaluation area.

[0229] The assessment process for long jump take-off violations may include: obtaining the third movement speed of the foot bones of the person's left and right feet based on the take-off frame in the video to be tested and the third frame before the take-off frame; and assessing the long jump take-off violation based on the third movement speed of the foot bones of the person's left and right feet and a preset violation speed threshold.

[0230] The assessment process for long jump interruption violations may include: performing human detection on the fourth frame after the first frame to obtain the human detection result of the fourth frame; if the human detection result of the fourth frame indicates that there is no one in the take-off area under the long jump assessment area, it is determined that the person has committed a long jump interruption violation.

[0231] In addition to the aforementioned violations such as stepping over the line, taking off, and interruption, standing violations and interference violations can also be assessed. This includes determining whether a person is standing in the wrong direction and whether there is interference during the long jump.

[0232] Furthermore, if any one or more violations occur during the long jump, or if there is interference, the system will directly determine that the long jump result is invalid and end the long jump evaluation process after the system announces "Result invalid".

[0233] The method provided in this invention determines the test video and scale coordinates within the long jump evaluation area. From each frame of the test video, a landing frame is determined. Based on the coordinates of the foot bone points of the person in the landing frame, the landing frame is segmented to obtain the coordinates of multiple foot contour points. Based on the coordinates of these multiple foot contour points and the scale coordinates, the long jump performance is evaluated. This overcomes the inaccuracy of traditional long jump performance calculations. By analyzing the state of the test personnel through human detection, posture estimation, and foot segmentation, the landing frame corresponding to the landing frame is determined, and the performance is evaluated accordingly, thus improving the accuracy of long jump performance calculation. Furthermore, the use of computer vision technology and a cyclic state machine for long jump evaluation also improves the efficiency of long jump evaluation and the teaching efficiency of physical education classes.

[0234] The long jump evaluation device provided by the present invention is described below. The long jump evaluation device described below can be referred to in correspondence with the long jump evaluation method described above.

[0235] Figure 14 This is a schematic diagram of the structure of the long jump evaluation device provided by the present invention, as shown below. Figure 14 As shown, the device includes:

[0236] Unit 1410 is used to determine the coordinates of the video to be tested and the scale lines in the long jump evaluation area;

[0237] The landing frame determination unit 1420 is used to determine the landing frame from each frame in the video to be tested;

[0238] The contour point determination unit 1430 is used to segment the landing frame into foot parts based on the coordinates of the foot bone points of the person in the landing frame, and obtain the coordinates of multiple foot contour points in the landing frame.

[0239] The performance evaluation unit 1440 is used to evaluate the long jump performance based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale line.

[0240] The long jump evaluation device provided by this invention determines the test video and scale line coordinates within the long jump evaluation area. From each frame of the test video, it identifies the landing frame and, based on the coordinates of the foot bone points of the person in the landing frame, segments the landing frame into multiple foot contour points, obtaining the coordinates of these points. Based on the coordinates of these foot contour points and the scale line coordinates, the long jump performance is evaluated. This overcomes the inaccuracy of traditional long jump performance calculations. By analyzing the tester's state through human detection, posture estimation, and foot segmentation, the device determines the corresponding landing frame and evaluates the performance accordingly, improving the accuracy of long jump performance calculation. Furthermore, the use of computer vision technology and a cyclic state machine for long jump evaluation also improves the efficiency of long jump evaluation and physical education teaching.

[0241] Based on the above embodiments, the performance evaluation unit 1440 is used for:

[0242] Based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale line, the scale interval where the multiple foot contour points are located is determined.

[0243] The person's long jump score is determined based on the upper or lower limit of the scale interval and the distance between the plurality of foot contour points and the upper or lower limit of the scale interval.

[0244] Based on the above embodiments, the performance evaluation unit 1440 is used for:

[0245] Curve fitting is performed on multiple foot contour points in the landing frame, and the foot contour curves obtained from the curve fitting are then sparsified.

[0246] Based on the coordinates of multiple foot contour points after sparsification, and the coordinates of the scale line, the scale interval in which the multiple foot contour points are located is determined.

[0247] Based on the above embodiments, the device further includes an attitude estimation unit, used for:

[0248] Based on the attitude estimation model, the attitude of the person in the landing frame is estimated to obtain the coordinates of the foot bone points of the person in the landing frame;

[0249] The posture estimation model is trained based on a first sample image, the region label of the long jump evaluation area in the first sample image, and the human skeleton point label, with the constraint that the person is in the long jump evaluation area for long jump evaluation.

[0250] Based on the above embodiments, the landing frame determination unit 1420 is used for:

[0251] Based on the human detection result of the first frame in the video to be tested, the state of the person is determined, and when the state is in the ready state, the first movement speed of the foot bone points of the person is obtained based on each frame in the video to be tested.

[0252] When the first moving speed indicates that the person is in a jumping state, the second moving speed of the person's foot bone points is obtained based on each frame in the video to be tested.

[0253] If the second moving speed is less than a preset landing speed threshold, it is determined that the person is in a landing state, and the landing frame corresponding to the landing state is determined from each frame in the video to be tested.

[0254] Based on the above embodiments, the device further includes a jump state verification unit, used for:

[0255] From each frame in the video to be tested, determine the jump frame corresponding to the jump state;

[0256] Based on the coordinates of the person's foot bones in the take-off frame and the coordinates of the person's foot bones in the second frame after the take-off frame in the video to be tested, the displacement of the person's foot bones is determined.

[0257] Based on the coordinates of the person's foot bones in the second frame, the displacement of the person's foot bones, and the coordinates of the take-off line in the long jump evaluation area, the take-off state is verified, and if the verification passes, the second movement speed of the person's foot bones is obtained based on each frame after the second frame in the video to be tested.

[0258] Based on the above embodiments, the device further includes a take-off state determination unit, used for:

[0259] Based on the first moving speed, determine the horizontal moving speed component and the vertical moving speed component of the person's foot bone points.

[0260] If the horizontal velocity component and the vertical velocity component are both greater than a preset soft velocity threshold, and the horizontal velocity component or the vertical velocity component is greater than a preset hard velocity threshold, then the person is determined to be in a jumping state.

[0261] The preset soft speed threshold is less than the preset hard speed threshold.

[0262] Based on the above embodiments, the device further includes a violation assessment unit, used for:

[0263] When the state is in the ready state, pose estimation is performed on the first frame to obtain the coordinates of the foot bone points of the person in the first frame;

[0264] Based on the coordinates of the foot bone points of the person in the first frame, the first frame is segmented into foot parts to obtain the coordinates of multiple foot contour points in the first frame.

[0265] Based on the coordinates of multiple foot contour points in the first frame and the coordinates of the take-off line under the long jump evaluation area, a violation assessment of stepping on the line in the long jump is performed.

[0266] Based on the above embodiments, the violation assessment unit is used for:

[0267] Based on the jump frame in the video to be tested, and the third frame before the jump frame, the third movement speed of the foot bone points of the person's left and right feet is obtained.

[0268] The long jump take-off violation is assessed based on the third movement speed of the foot bones of the person's left and right feet, and a preset violation speed threshold.

[0269] Based on the above embodiments, at least one of the preset landing speed threshold, the preset soft speed threshold, and the preset hard speed threshold is determined based on the coordinates of the scale line under the long jump evaluation area.

[0270] Figure 15 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 15 As shown, the electronic device may include a processor 1510, a communications interface 1520, a memory 1530, and a communication bus 1540, wherein the processor 1510, the communications interface 1520, and the memory 1530 communicate with each other via the communication bus 1540. The processor 1510 can call logical instructions in the memory 1530 to execute a long jump evaluation method, which includes: determining the test video and scale line coordinates under the long jump evaluation area; determining the landing frame from each frame in the test video; segmenting the landing frame based on the coordinates of the foot bone points of the person in the landing frame to obtain the coordinates of multiple foot contour points in the landing frame; and evaluating the long jump performance based on the coordinates of the multiple foot contour points in the landing frame and the scale line coordinates.

[0271] Furthermore, the logical instructions in the aforementioned memory 1530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0272] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the long jump evaluation method provided by the above methods, the method comprising: determining a test video and scale line coordinates under a long jump evaluation area; determining a landing frame from each frame in the test video; segmenting the landing frame into foot parts based on the coordinates of the foot bone points of the person in the landing frame to obtain the coordinates of multiple foot contour points in the landing frame; and evaluating the long jump performance based on the coordinates of the multiple foot contour points in the landing frame and the scale line coordinates.

[0273] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the long jump evaluation method provided by the above methods. The method includes: determining a test video and scale line coordinates under a long jump evaluation area; determining a landing frame from each frame in the test video; segmenting the landing frame into foot parts based on the coordinates of the foot bone points of the person in the landing frame to obtain the coordinates of multiple foot contour points in the landing frame; and evaluating the long jump performance based on the coordinates of the multiple foot contour points in the landing frame and the scale line coordinates.

[0274] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0275] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0276] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the long jump, characterized in that, include: Determine the coordinates of the video to be tested and the scale lines within the long jump evaluation area; Determine the landing frame from each frame in the video to be tested; Based on the coordinates of the foot bone points of the person in the landing frame, the landing frame is segmented into foot parts to obtain the coordinates of multiple foot contour points in the landing frame. The long jump performance is evaluated based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines. Determining the landing frame from each frame in the video to be tested includes: When the person in the video to be tested is in a ready state, the first movement speed of the person's foot bone points is obtained based on each frame of the video to be tested. When the first moving speed indicates that the person is in a jumping state, the take-off frame corresponding to the jumping state is determined from each frame in the video to be tested; Based on the coordinates of the person's foot bones in the take-off frame and the coordinates of the person's foot bones in the second frame after the take-off frame in the video to be tested, the displacement of the person's foot bones is determined. Based on the coordinates of the person's foot bones in the second frame, the displacement of the person's foot bones, and the coordinates of the take-off line under the long jump evaluation area, the take-off state is verified, and if the verification is successful, the second movement speed of the person's foot bones is obtained based on each frame after the second frame in the video to be tested. Based on the second moving speed, the landing frame corresponding to the landing state is determined from each frame in the video to be tested.

2. The long jump evaluation method according to claim 1, characterized in that, The evaluation of the long jump performance based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale lines includes: Based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale line, the scale interval where the multiple foot contour points are located is determined. The person's long jump score is determined based on the upper or lower limit of the scale interval and the distance between the plurality of foot contour points and the upper or lower limit of the scale interval.

3. The long jump evaluation method according to claim 2, characterized in that, Determining the scale interval where the multiple foot contour points are located based on the coordinates of the multiple foot contour points in the landing frame and the coordinates of the scale line includes: Curve fitting is performed on multiple foot contour points in the landing frame, and the foot contour curves obtained from the curve fitting are then sparsified. Based on the coordinates of multiple foot contour points after sparsification, and the coordinates of the scale line, the scale interval in which the multiple foot contour points are located is determined.

4. The long jump evaluation method according to claim 1, characterized in that, The coordinates of the foot bone points are determined based on the following steps: Based on the attitude estimation model, the attitude of the person in the landing frame is estimated to obtain the coordinates of the foot bone points of the person in the landing frame; The posture estimation model is trained based on a first sample image, the region label of the long jump evaluation area in the first sample image, and the human skeleton point label, with the constraint that the person is in the long jump evaluation area for long jump evaluation.

5. The long jump evaluation method according to any one of claims 1 to 4, characterized in that, The state of the people in the video to be tested is determined based on the following steps: Based on the human detection results of the first frame in the video to be tested, the state of the person is determined; The step of determining the landing frame corresponding to the landing state from each frame in the video under test based on the second moving speed includes: If the second moving speed is less than a preset landing speed threshold, it is determined that the person is in a landing state, and the landing frame corresponding to the landing state is determined from each frame in the video to be tested.

6. The long jump evaluation method according to claim 5, characterized in that, The take-off state is determined based on the following steps: Based on the first moving speed, determine the horizontal moving speed component and the vertical moving speed component of the person's foot bone points. If the horizontal velocity component and the vertical velocity component are both greater than a preset soft velocity threshold, and the horizontal velocity component or the vertical velocity component is greater than a preset hard velocity threshold, then the person is determined to be in a jumping state. The preset soft speed threshold is less than the preset hard speed threshold.

7. The long jump evaluation method according to claim 5, characterized in that, Also includes: When the state is in the ready state, pose estimation is performed on the first frame to obtain the coordinates of the foot bone points of the person in the first frame; Based on the coordinates of the foot bone points of the person in the first frame, the first frame is segmented into foot parts to obtain the coordinates of multiple foot contour points in the first frame. Based on the coordinates of multiple foot contour points in the first frame and the coordinates of the take-off line under the long jump evaluation area, a violation assessment of stepping on the line in the long jump is performed.

8. The long jump evaluation method according to claim 5, characterized in that, Also includes: Based on the jump frame in the video to be tested, and the third frame before the jump frame, the third movement speed of the foot bone points of the person's left and right feet is obtained. The long jump take-off violation is assessed based on the third movement speed of the foot bones of the person's left and right feet, and a preset violation speed threshold.

9. The long jump evaluation method according to claim 6, characterized in that, At least one of the preset landing speed threshold, the preset soft speed threshold, and the preset hard speed threshold is determined based on the coordinates of the scale line under the long jump evaluation area.

10. A long jump evaluation device, characterized in that, include: The determination unit is used to determine the coordinates of the video to be tested and the scale lines within the long jump evaluation area; The landing frame determination unit is used to determine the landing frame from each frame in the video to be tested; The contour point determination unit is used to segment the landing frame into foot sections based on the coordinates of the foot bone points of the person in the landing frame, and obtain the coordinates of multiple foot contour points in the landing frame. The performance evaluation unit is used to evaluate the long jump performance based on the coordinates of multiple foot contour points in the landing frame and the coordinates of the scale line. Determining the landing frame from each frame in the video to be tested includes: When the person in the video to be tested is in a ready state, the first movement speed of the person's foot bone points is obtained based on each frame of the video to be tested. When the first moving speed indicates that the person is in a jumping state, the take-off frame corresponding to the jumping state is determined from each frame in the video to be tested; Based on the coordinates of the person's foot bones in the take-off frame and the coordinates of the person's foot bones in the second frame after the take-off frame in the video to be tested, the displacement of the person's foot bones is determined. Based on the coordinates of the person's foot bones in the second frame, the displacement of the person's foot bones, and the coordinates of the take-off line under the long jump evaluation area, the take-off state is verified, and if the verification is successful, the second movement speed of the person's foot bones is obtained based on each frame after the second frame in the video to be tested. Based on the second moving speed, the landing frame corresponding to the landing state is determined from each frame in the video to be tested.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the long jump evaluation method as described in any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the long jump evaluation method as described in any one of claims 1 to 9.

Images