Multi-person sporting event counting method and system based on multi-key-point detection, and medium

Abstract

The invention discloses a multi-person sports item counting method and system based on multi-key-point detection and a medium, and the method comprises the steps: obtaining video stream data of a detection site in real time through a camera; the method comprises the following steps: preprocessing a video frame in a video stream to obtain a final video frame for key point detection, detecting a bounding box position of a to-be-detected person in the video frame, intercepting an image area for key point detection, tracking the person in the video frame, detecting coordinate values of 33 key points in a human body, and performing key point detection. Counting of corresponding items is completed for different sporting events. The method can accurately identify the information of each key point of the to-be-detected person under different resolutions, different illumination intensities and different field backgrounds, accurately completes the counting work of corresponding items, has the advantages of high accuracy, good generalization and strong expansibility, and can be widely applied to the work of counting, timing and the like in the related fields of sports.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a multi-person sports event counting method, device and medium based on multi-key point detection. Background Art

[0002] Sports counting (timing) has a wide range of uses in various scenarios, such as competitions, sports exams, etc. The current counting methods can be roughly divided into three categories. The first is the traditional manual counting method, which is time-consuming and labor-intensive, and will be affected by the subjective judgment of the counting personnel; the second is the use of multiple sensor-type physical devices to assist in counting, which can greatly save labor costs, but still requires the purchase of a large number of physical equipment, which is costly; the last is a pure visual solution based on deep neural networks, which only requires physical equipment and computing equipment that can record the video of the detection site to achieve a high-accuracy and robust counting solution.

[0003] There are some problems with the current counting schemes based on computer vision, such as most of them only support single-person detection, or the number of detected key points of the human body is small, which cannot meet the requirements of project counting. For key point detection of the human body, there are two mainstream methods. One is the top-down method, that is, first detect the bounding box of the human body, and then detect the key points of the person in the detection box. This method can realize the key point detection of multiple people, but the number of key points detected by the existing schemes is small, such as YOLO-POSE has only 17, which cannot meet the requirements of fine algorithms, and its model structure has weak perception of small objects and is difficult to be competent for the task of multi-key point detection; the other is the bottom-up method, that is, directly identify the key points of the person in the image, and select the key points with the highest probability to form the human body structure. This method can realize a large number of key point detection, but cannot realize the detection of multiple people, and the detection of multiple people requires the determination of the identity information of each person to be detected, that is, each person should have a unique and unchanged ID, so it is difficult to achieve accurate multi-person detection and counting by relying solely on the key point detection algorithm. Summary of the invention

[0004] Technical problem to be solved by the present invention: In view of the above-mentioned problems in the prior art, a method, system and medium for counting multi-person sports events based on multi-key point detection are provided. The present invention can accurately complete the counting (timing) of corresponding multi-person sports events under different resolutions, different venue backgrounds and different light intensities. It has the advantages of high accuracy and good generalization, and can be widely used in sports events, sports examinations and other scenarios.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A multi-person sports event counting method based on multi-key point detection, the implementation steps include:

[0007] 1) Get the video stream data of the venue.

[0008] 2) Preprocessing the video frames in the video stream to obtain the final video frames for key point detection, wherein the preprocessing of the video frames includes sampling the video frames, pre-storing the video frames and possible scaling operations on the video frames.

[0009] 3) Detect the bounding box position of the person to be detected in the video frame and intercept the image area for key point detection. The bounding box position is detected using a target detection model based on YOLO V8. This model has a fast inference speed and can meet the requirements of real-time detection. For the identified area of ​​the person to be detected, on the one hand, it is necessary to intercept the area for subsequent key point detection, and on the other hand, it is necessary to use it for subsequent target tracking tasks; after completing the target detection, it is necessary to track the person to be detected. This part uses the DeepSort algorithm to assign a unique and unchanging ID to each person to be detected to facilitate subsequent counting operations.

[0010] 4) Detect the coordinates of 33 key points in the human body, including 11 on the head, 10 on the arms (on both sides), 4 on the torso, 2 on the legs (on both sides), and 6 on the feet (on both sides), using the posture detection module of mediapipe;

[0011] 5) Count the number of completed items for different sports. Take the sit-up count as an example. By defining four states of the person to be tested, namely, the ready state, the supine state, the sit-up state, and the end state, and using corresponding rules to determine the transition between states, the person to be tested is deemed to have completed one sit-up when he or she has completed the transition from the supine state to the sit-up state to the supine state.

[0012] A multi-person sports counting system based on multi-key point detection comprises a computer device which is programmed or configured to execute the steps of any one of the multi-person sports counting methods based on multi-key point detection described above, or a computer program programmed or configured to execute any one of the multi-person sports counting methods based on multi-key point detection described above is stored on the storage medium of the computer device.

[0013] A computer-readable storage medium stores a computer program programmed or configured to execute any one of the above-mentioned methods for counting multi-person sports events based on multi-key point detection.

[0014] Compared with the prior art, the present invention has the following advantages: preprocessing is performed on the video stream to prevent the problem of video stream buffer overflow when the processing speed is insufficient; multiple key point detection of multiple people is realized, and the precision and accuracy of the counting algorithm are improved while taking into account the efficiency of key point detection; at the same time, due to the use of DeepSort for target tracking, tracking failures caused by instantaneous occlusion can be effectively avoided, and the counting of multiple people in sports events can be accurately completed in scenarios with different venue backgrounds, light intensities, and numbers of people. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Schematic diagram of the basic flow of the method of the embodiment of the present invention.

[0016] Figure 2 Schematic diagram of the basic flow of the tracking algorithm according to an embodiment of the present invention.

[0017] Figure 3 This is the network architecture diagram of the DeepSort algorithm feature extraction network.

[0018] Figure 4 This is a key point annotation diagram for key point detection of a human body according to an embodiment of the present invention.

[0019] Figure 5 This is a state transition diagram of the sit-up counting algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] like Figure 1 As shown, the implementation steps of the multi-person sports event counting method based on multi-key point detection in this embodiment include:

[0021] 1. Obtain the video stream data of the site, which is usually obtained by a camera device that supports transmitting video streams via wired or wireless means, usually a camera device that supports RTSP, RTMP or encoded video streams. At the same time, this method also supports local camera devices, and then counts (times) the stored videos to be detected accordingly.

[0022] 2. Preprocess the video frames in the video stream to obtain the final video frames for key point detection. The detailed steps of the preprocessing include:

[0023] 2.1) The video frames in the original video stream are uniformly sampled. For video streams with a FPS higher than 30, the frame rate after sampling is 20 frames per second.

[0024] 2.2) Create a data structure called FramePool, whose size is determined by the hardware conditions such as the memory of the specific computing device. The video frames sampled in 2.1) are stored in the FramePool first. When the capacity of FramePool is insufficient and a new video frame arrives, the first video frame in FramePool is deleted first, and the newly arrived video frame is stored in FramePool. At the same time, the subsequent process will take the first video frame in FramePool for processing. The above process is an asynchronous operation, and there is a corresponding lock mechanism to ensure the correctness of FramePool access.

[0025] 2.3) For video frames sampled from FramePool, if their resolution is higher than 1280×720, they are scaled to 1280×720 resolution to facilitate subsequent processing. After this preprocessing, the problem of video stream buffer overflow caused by insufficient computing power of the computing device is effectively avoided.

[0026] 3. Detect the bounding box position of the person to be detected in the video frame and capture the image area for key point detection. At the same time, track the person to be detected. The specific implementation steps are as follows:

[0027] 3.1) Use the target detection model based on YOLO V8 to detect the persons to be detected in the video frame. The BackBone of this model uses the C2f module as the basic building block. Specifically, it consists of 5 convolution modules and 4 c2f modules. Compared with the previous generation C3 module, it has fewer parameters and better feature extraction capabilities, which is more in line with the needs of real-time detection of video stream data; its Neck uses multi-scale feature fusion technology to fuse feature maps from different stages of Backbone to enhance feature representation capabilities. This part plays an important role in feature extraction and feature fusion; and the Head part is responsible for the final target detection and classification tasks, including a detection head and a classification head. The detection head contains a series of convolutional layers and deconvolution layers to generate detection results; the classification head uses global average pooling to classify each feature map. Through the bounding box information obtained, the key point to be detected area in the image, that is, the area containing the person to be detected, is cropped.

[0028] 3.2) Based on the bounding box information obtained in 3.1) and the captured key point detection area image, use the DeepSort algorithm to track the person to be detected in the video frame. The specific implementation steps are as follows: Figure 2 As shown: The image captured in 3.1) is subjected to feature extraction, and the feature will be used for cascade matching later. The structure of the feature extraction network is as follows Figure 3As shown in the figure: the pre-convolution layer consists of 2D convolution and BN layers, which are processed by the subsequent convolution layer composed of basic blocks after ReLU and maximum pooling operations. The basic block consists of 2 2D convolution layers connected to BN, and the activation function uses ReLU. In particular, the basic block provides an optional downsampling operation. The backbone network consists of 8 basic blocks and an average pooling layer, where every two basic blocks form a group. If the group of basic blocks chooses to downsample, only the first basic block in the group is downsampled, that is, the 3rd, 5th, and 7th basic blocks in the current structure are downsampled. The classifier consists of a linear layer and a BN layer. After ReLU activation and Dropout operations, it is connected to a linear layer for final classification to obtain the result. The extracted features will be used for subsequent cascade matching.

[0029] The trajectory of the bounding box, i.e., trace, has two states, namely, confirmed state and unconfirmed state. The unconfirmed state trajectory can be converted into the confirmed state trajectory under certain conditions. In the subsequent matching process, there are different processing rules for the unconfirmed state trajectory and the confirmed state trajectory. Each existing trajectory will be predicted by the Kalman filter (hereinafter referred to as KF) for its subsequent trajectory.

[0030] For a frame where a person's bounding box is detected for the first time, the bounding box is initialized as the original trajectory. All trajectories are in an unconfirmed state when they are first generated. For unconfirmed trajectories, their KF predicted trajectories will be IOU matched with the detection boxes of subsequent video frames. The match will produce three matching results through the Hungarian algorithm. If the trajectory is not successfully matched, the trajectory will be deleted (since the trajectory is in an unconfirmed state, it can be deleted directly). If the detection box is not matched successfully, the detection box will be initialized as a new trajectory. If the two match successfully, the trajectory will be updated according to the position of the detection box.

[0031] If a trajectory matches a detected bounding box for multiple consecutive times, the trajectory is transformed into a confirmed state. The number of consecutive matches is a variable threshold, usually set to 3. For confirmed trajectories, its KF predicted trajectory will be cascade matched with the detection box in the subsequent video frame. The cascade match will use the features extracted by the deep neural network mentioned above. Similarly, the match will have two results. If the detection box does not match or the trajectory does not match, a subsequent IOU match will be performed. If the match is successful, the trajectory information is updated according to the detection box position. For confirmed trajectories, if the IOU match fails, the trajectory will not be deleted directly, but will be deleted after multiple consecutive matching failures. The number of consecutive multiple failures is a variable threshold, usually set to 30.

[0032] 4. Detect the coordinate values ​​of 33 key points on the human body, see the attached Figure 4, including 11 heads, 10 arms (on both sides), 4 torsos, 2 legs (on both sides), and 6 feet (on both sides); the specific key point parameters are 0-nose、1-left eye(inner)、2-left eye、3-left eye(outer)、4-right eye(inner)、5-right eye、6-righteye(outer)、7-left ear、8-right ear、9-mouth(left)、10-mouth(right)、11-leftshoulder、12-right shoulder、13-left elbow、14-right elbow、15-left wrist、16-right wrist、17-leftpinky、18-rightpinky、19-left index、20-right index、21-leftthumb、22-right thumb、23-lefthip、24-righthip、25-left knee、26-right knee、27-left ankle、28-right ankle、29-leftheel、30-right heel、31-left foot index, 32-right foot index. This key point detection model uses the pose module of mediapipe. Compared with mainstream models such as YOLO-POSE, this module recognizes more key points and has a faster recognition speed. The subdivision of the key points of the human foot gives it a unique advantage in processing some projects that require fine foot judgment. Its disadvantage is that due to the use of a bottom-up model architecture, that is, identifying all possible human key points in the image and selecting several key points with the highest confidence to form the human skeletal structure, the model only supports single-person human key point detection. When there are multiple people to be detected in the image, the model only outputs the key point of the "most credible" person to be detected. Therefore, in response to the above problems, this method uses this model in combination with target detection and target tracking to achieve multi-key point and fast multi-person skeletal key point recognition.

[0033] 5. For different sports, the corresponding counting algorithms are used to complete the counting of the corresponding sports. The following takes the counting of sit-ups as an example. Figure 5 The specific steps of this embodiment are as follows:

[0034] 5.1) Define the state of the person to be tested. There are four states in total, namely, the preparation state, the supine state, the sitting state, and the end state. The specific definitions of each state are as follows. Preparation state: The person appears in the picture and is detected by the target detection model, but has not started the sit-up exercise; Supine state: The person lies in the test position, with the torso horizontal, the shoulder blades touching the ground, the legs naturally bent, the hands placed on both sides of the body close to the ground, and the soles of the feet touching the ground; Sit-up state: The tester relies on the strength of the abdomen to lift the upper body to a certain angle, and at the same time moves the fingers forward to the standard line position. During the process, the feet and buttocks are prohibited from leaving the ground; End state: When the person to be tested stands up from the supine state or the sitting state, it enters the end state, and no counting judgment is performed after entering the end state.

[0035] 5.2) Ready state determination: After the person to be detected appears in the video screen and is detected by the aforementioned target detection model, the tracking algorithm assigns the person a unique ID and the person enters the ready state, at which time the key points of the human skeleton of the person to be detected begin to be identified.

[0036] 5.3) For the detected key points, confidence screening will be performed before subsequent use. Those with confidence less than the set threshold will not be used. For data with bilateral key points, if both sides of the key points are greater than the set threshold, the average value will be taken to reduce misjudgment caused by errors in the key point detection model.

[0037] 5.4) Supine state determination: The person to be tested should be in a horizontal lying state rather than an upright state. At this time, the x-direction coordinate difference between the key points of the head and the key points of the feet should be greater than the y-direction coordinate difference, and the x-direction coordinate difference between the key points of the hip and the key points of the feet should be greater than the y-direction coordinate difference. The head key point uses the coordinates of the key point of the nose position, and the foot key point uses the average coordinates of the three key points of the foot, namely the ankle, heel, and toe. At the same time, it is necessary to ensure that the person lies horizontally on the ground with the shoulder blades close to the ground, that is, the angle formed by the three key points of the head, hip, and foot is close to 180 degrees; the legs are naturally bent, requiring the angle formed by the three key points of the hip, knee, and foot to be close to 90 degrees; the feet touch the ground, that is, while meeting the requirements of lying horizontally, the angles of the three key points of the toes, heels, and hips are close to 180 degrees. Among them, the method used to calculate the angle based on the coordinates of the three points is:

[0038]

[0039] Note: x2 and y2 are the coordinates of the key points of the vertex of the angle.

[0040] 5.5) Sitting-up position determination: After the person being tested enters the supine position, the determination of whether he / she has entered the sitting-up position begins. First, detect whether the frames in the sitting-up process show the phenomenon of hip lifting, foot lifting, etc. If so, the count will be invalidated. When the character enters the supine state, the coordinates of the key points of the character's hip and foot will be recorded. According to the changes in the y-direction coordinates of the character's hip and foot during the standing-up process, it is judged whether the above-mentioned errors occur. Among them, foot lifting includes foot lifting and foot lifting. For foot lifting and foot lifting, it is only necessary to compare the average value of the three key points of the foot, namely the ankle, heel, and toe, with the average value of the foot key points recorded initially, so as to judge whether the above-mentioned two errors occur at the same time; if no such errors occur, detect the angle of the character's standing up, and judge the angle of the character's standing up according to the angle formed by the three key points of the shoulder, hip, and foot. When the angle meets a certain threshold, it is considered to have met the requirements, which is usually set to 150 degrees, that is, the angle between the upper body of the person to be tested and the ground is greater than 30 degrees; finally, detect whether the hand moves to the standard line, and judge whether the hand has crossed the line according to the finger coordinates and the standard line coordinate position.

[0041] 5.6) End state determination: If the current state of the person to be detected is in a supine state or a sitting state, and the current person no longer satisfies the horizontal lying state described in 5.4), that is, the person to be detected has stood up, then the current person enters the end state, and no counting determination is performed after entering the end state.

[0042] 5.7) When the person being tested goes from the supine state to the sit-up state and then back to the supine state, it is deemed that he has completed one sit-up action, and the sit-up count of the person is increased accordingly. It should be noted that the supine state can be obtained by changing from the ready state and the sit-up state, the sit-up state can only be obtained by changing from the supine state, and the end state can be obtained by changing from the supine state and the sit-up state.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Other modifications or equivalent substitutions made to the technical solution of the present invention by ordinary technicians in the field should be included in the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solution of the present invention.

Claims

1. A multi-person sports event counting method based on multi-key point detection, characterized in that: include: 1) Obtain the video stream data of the venue; 2) Preprocessing the video frames in the video stream to obtain the final video frames for key point detection; 3) Detect the bounding box position of the person to be detected in the video frame and capture the image area for key point detection, while tracking the person to be detected; 4) Detect the coordinate values ​​of 33 key points on the human body, including 11 on the head, 10 on the arms (on both sides), 4 on the torso, 2 on the legs (on both sides), and 6 on the feet (on both sides); 5) Complete the counting of corresponding items for different sports.

2. The method for counting multi-person sports events based on multi-key point detection according to claim 1 is characterized in that: Step 2) includes: 2.1) The video frames are uniformly sampled. The sampling step is set according to the actual computing device and the frame rate of the original video stream. Usually, for a video stream with a frame rate greater than or equal to 30, the sampled video frames are 20 frames per second; 2.2) Establish a FramePool, which is used to store sampled video frames. Set the size of the FramePool according to the actual computing device. When the storage is full and new frames enter, discard the first frame that enters the FramePool; 2.3) The processing thread will continuously obtain the earliest frames from the FramePool, resize them, and then perform subsequent processing. Specifically, video frames with a resolution greater than 1280×720 will be scaled to 1280×720, and frames with a resolution too small will not be processed. After this preprocessing, the problem of video stream cache overflow caused by insufficient computing power of the computing device can be effectively avoided.

3. The method for counting multi-person sports events based on multi-key point detection according to claim 1 is characterized in that: Step 3) includes using a pre-trained target detection model to detect the video frame to determine the position of the person to be detected, and then using a tracking algorithm to track the detected person, that is, determining the bounding box and unique ID of each person to be detected in the video, where the unique ID is assigned by this system and is only used for target tracking.

4. The method for counting multi-person sports events based on multi-key point detection according to claim 3 is characterized in that: The pre-trained target detection model uses a deep convolutional neural network model based on YOLO V8 to perform target detection on the video frames obtained in step 2.3) of claim 2. The BackBone of the model uses the C2f module as the basic building block. Compared with the previous generation C3 module, it has fewer parameters and better feature extraction capabilities, and is more in line with the needs of real-time detection of video stream data; Its Neck adopts multi-scale feature fusion technology to fuse feature maps from different stages of Backbone to enhance feature representation capabilities; the Head part is responsible for the final target detection and classification tasks, including a detection head and a classification head; the detection head contains a series of convolutional layers and deconvolution layers to generate detection results; the classification head uses global average pooling to classify each feature map; through the obtained bounding box information, the key point area to be detected in the image, that is, the area containing the person to be detected, is cropped.

5. The method for counting multi-person sports events based on multi-key point detection according to claim 3 is characterized in that: To track the person to be detected, use the DeepSort algorithm. The steps include: s1) extracting features from the image captured in claim 4, which will be used for cascade matching later; s2) setting the person detection frame in the first processed frame containing a plurality of person detection frames as the initial trajectory; s3) Use Kalman filtering to predict the trajectory of the person. The trajectory is divided into confirmed and unconfirmed states. For the unconfirmed trajectory box, the predicted person bounding box is matched with the subsequently detected ground truth by IOU, and the linear matching result is obtained by the Hungarian algorithm. If the predicted trajectory does not match, the trajectory is deleted. If the detected box does not match, a new trajectory is generated, indicating that a new person is detected. If it matches, the trajectory information is updated. For the confirmed trajectory, cascade matching is performed with the detected bounding box. In this step, the appearance features and trajectory information extracted in s1) are used at the same time. If the match is successful, the trajectory is updated. The trajectory that fails to match will continue to perform subsequent IOU matching. s4) For a trajectory in an unconfirmed state, if it is successfully matched several times in a row, it can be converted to a confirmed state. s5) For a confirmed trajectory, if the number of IOU matching failures is greater than the set threshold, the trajectory is deleted; otherwise, the trajectory is put back into the trajectory set and waits for the next match.

6. The method for counting multi-person sports events based on multi-key point detection according to claim 1, characterized in that: The coordinate values ​​of 33 key points of the human body are detected using the posture detection module of mediapipe. The specific key points detected are 11 on the head, including eyes, nose, mouth, and ears; 10 on the arms (on both sides), including elbows, wrists, and fingers; 4 on the torso, including shoulders and hips; 2 on the legs (on both sides), including knees; 6 on the feet (on both sides), including ankles, heels, and toes. The acquired coordinates are two-dimensional coordinates and the corresponding visibility information, that is, the coordinate format of a single key point is [x, y, visibility].

7. The method for counting multi-person sports events based on multi-key point detection according to claim 1 is characterized in that: For sit-ups, the steps are as follows: c1) Define the status of the person to be tested: Ready state: the person appears in the screen but has not started to do sit-ups; Supine state: the person lies in the test position, with the torso horizontal, shoulder blades touching the ground, legs naturally bent, hands placed on both sides of the body close to the ground, and soles of the feet touching the ground; Sit-up state: the tester lifts the upper body to a certain angle with the strength of the abdomen, and the fingers move forward to the standard line position. During the process, the feet and buttocks are prohibited from leaving the ground; End state: when the person to be tested stands up from the supine state or the sit-up state, it enters the end state, and no counting judgment is performed after entering the end state; c2) Ready state determination: after the person to be detected appears in the video screen and is detected by the target detection model described in claim 3, the tracking algorithm assigns the person a unique ID, and the person enters the ready state; c3) Supine state determination: The person to be tested should be in a horizontal lying state rather than an upright state. At this time, the x-direction coordinate difference between the key points of the head and the foot should be greater than the y-direction coordinate difference, and the x-direction coordinate difference between the key points of the hip and the foot should be greater than the y-direction coordinate difference; at the same time, it is necessary to ensure that the person lies horizontally on the ground with the shoulder blade close to the ground, that is, the angle formed by the three key points of the head, hip and foot is close to 180 degrees; the legs are naturally bent, and the angle formed by the three key points of the hip, knee and foot is close to 90 degrees; the feet touch the ground, that is, the angles of the three key points of the toes, heels and hips are close to 180 degrees while meeting the requirements of lying horizontally; c4) Sitting up state judgment: After the person to be tested enters the supine state, it starts to judge whether he / she enters the sitting up state. First, it is detected whether the frame in the sitting up process shows the phenomenon of hip lifting, foot lifting, etc. If so, the count will be invalidated. When the person enters the supine state, the coordinates of the key points of the person's hip and foot will be recorded. According to the change of the y-direction coordinates of the person's hip and foot during the standing up process, it is judged whether the above-mentioned errors occur; if no such errors occur, the person's standing up angle is detected, and the standing up angle of the person is judged according to the angle formed by the three key points of the shoulder, hip and foot. When the angle meets a certain threshold, it is considered that the requirements are met; finally, it is detected whether the hand moves to the standard line, and it is judged whether the hand has crossed the line according to the coordinates of the finger and the coordinates of the standard line; c5) End state determination: If the current state of the person to be detected is in a supine state or a sitting state, and the current person no longer satisfies the horizontal lying state described in c3), the current person enters the end state, and no counting determination is performed after entering the end state; c6) When the person being tested moves from the supine position to the sit-up position and then back to the supine position, it is deemed that he has completed one sit-up action, and the person's sit-up count is increased accordingly.

8. A multi-player sports event counting system based on multi-key point detection, comprising a computer device, characterized in that: The computer device is programmed or configured to execute the steps of the method for counting multi-person sports events based on multi-key point detection as described in any one of claims 1 to 7, or the storage medium of the computer device stores a computer program that is programmed or configured to execute the method for counting multi-person sports events based on multi-key point detection as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program programmed or configured to execute the multi-person sports event counting method based on multi-key point detection as described in any one of claims 1 to 7.

Images