Multi-person shadowboxing motion image recognition scoring method and storage medium

Abstract

The invention relates to the technical field of computer vision, in particular to a multi-person shadowboxing motion image recognition scoring method and a storage medium. The scoring method comprises the following steps: synchronously shooting N paths of global images of multi-person shadowboxing movement in a target area from N different angles, and tracking three-dimensional coordinates of UWB tags worn by participants at shooting moments by using a UWB positioning technology; cutting out N paths of local images corresponding to each participant from the N paths of global images according to the three-dimensional coordinates of the UWB tag worn by each participant at each shooting moment; and detecting human body key point coordinates of the participant in each local image by using a pre-trained key point detection model, further generating N key point coordinate time sequences of the participant, respectively inputting the N key point coordinate time sequences into N pre-trained scoring models, obtaining N scores of the participant, and averaging the N scores to obtain a final score of the participant. According to the method, the scoring problem in multi-person shadowboxing sports is solved.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a multi-person Tai Chi movement image recognition and scoring method and a storage medium. Background Art

[0002] With the acceleration of modern life and the increasing social pressure, people are paying more and more attention to health issues. In this context, traditional fitness exercises, especially Tai Chi, have gradually gained popularity among the general public due to their unique health-preserving effects and slow-paced exercise methods. Tai Chi can not only enhance physical fitness and improve body coordination, but also has the function of self-cultivation and self-cultivation, and it also plays an important role in promoting mental health.

[0003] In the process of scoring Tai Chi, the traditional method mainly relies on manual work. This method is often influenced by the subjective influence of coaches or referees and lacks unified and objective judging standards. In recent years, with the breakthrough development of computer vision technology, image recognition and processing can be used to capture and analyze the movements and postures of athletes, thus providing an objective basis for sports scoring and improving the accuracy and fairness of scoring.

[0004] However, most of the existing technical solutions are designed for single-player sports. How to use computer vision technology to accurately identify and score the movements of each athlete in multi-player Tai Chi sports has become a problem that needs to be solved urgently. Summary of the invention

[0005] In order to solve the above problems in the prior art, the present invention proposes a multi-person Tai Chi exercise image recognition and scoring method, electronic equipment and storage medium, which solves the scoring problem in multi-person Tai Chi exercise.

[0006] A first aspect of the present invention provides a method for identifying and scoring multi-person Tai Chi sports images, the method comprising: N cameras are used to synchronously capture N global images of multiple people doing Tai Chi in the target area from N different angles, and UWB (Ultra Wide Band) positioning technology is used to track the three-dimensional coordinates of the UWB tags worn by each participant at each shooting moment; According to the three-dimensional coordinates of the UWB tags worn by each participant at each shooting moment, N local images corresponding to each participant are cropped from the N global images; According to the N local images corresponding to each participant, the coordinates of the human body key points of the participant in each local image are detected using a pre-trained key point detection model, thereby generating a time series of N key point coordinates of the participant; Inputting the N key point coordinate time series of each participant into N pre-trained scoring models respectively to obtain N scores of the participant; averaging the N scores of each participant to obtain a final score of the participant; Wherein, N is a preset value.

[0007] Preferably, the target area is a rectangular sports field, and the outer frame is drawn in a striking color; The N cameras are pre-fixed above or diagonally above the target area, and each camera is located higher than all participants to avoid overlapping portraits, and the shooting range of each camera is sufficient to cover the target area; Each participant wears a UWB positioning tag on his chest, and each UWB positioning tag has an ID number, which serves as the ID of the corresponding participant; Four UWB positioning base stations are set directly above the four vertices of the target area and are located higher than all participants; The N cameras, all UWB positioning tags and the four UWB positioning base stations are pre-calibrated at a unified time.

[0008] Preferably, the step of “cutting out N local images corresponding to each participant from the N global images according to the three-dimensional coordinates of the UWB tags worn by each participant at each shooting moment” comprises: Preprocessing the N global images to obtain N preprocessed images; Calculate the two-dimensional projection coordinates of the UWB tag of each participant in each image based on the N pre-processed images and the three-dimensional coordinates of the UWB tag of each participant at each shooting time; According to the N pre-processed images and the two-dimensional projection coordinates of each participant, using a pre-trained human figure detection model, detecting the position of the rectangular frame of each participant in each pre-processed image; The N pre-processed images are cropped according to the positions of the rectangular frames, and the cropped rectangular images are scaled to a preset first size to obtain N local images corresponding to each participant.

[0009] Preferably, the step of "calculating the two-dimensional projection coordinates of the UWB tag of each participant in each image according to the N pre-processed images and the three-dimensional coordinates of the UWB tag of each participant at each shooting moment" includes: Calculate the vertical projection coordinates of the UWB tag of each participant in the plane where the target area is located at each shooting time according to the positions of the four UWB positioning base stations and the three-dimensional coordinates of the UWB tag of each participant at each shooting time; OpenCV (Open Source Computer Vision Library, an open source computer vision and machine learning software library) is used to detect the coordinates of the four vertices of the image corresponding to the target area in each image; According to the projection coordinates of each participant's UWB tag in the target area at each shooting moment and the four vertex coordinates of the image corresponding to the target area in each image, the two-dimensional projection coordinates of each participant's UWB tag in the image are calculated.

[0010] Preferably, the step of "detecting the coordinates of the human body key points of the participant in each local image using a pre-trained key point detection model according to the N local images corresponding to each participant, and then generating a time series of N key point coordinates of the participant" includes: Select one image from the first path of partial images of a participant at a time and input it into the pre-trained key point detection model to obtain the two-dimensional human body key point coordinates of the participant on each image in the first path of partial images, and then generate the first key point coordinate time series of the participant; Select one image from the second group of partial images of the participant at a time and input it into the key point detection model to obtain the two-dimensional coordinates of the human body key points of the participant on each image in the second group of partial images, thereby generating a second key point coordinate time series of the participant; The process is repeated until the Nth key point coordinate time series of the participant corresponding to the Nth local image is generated.

[0011] Preferably, the training step of the human figure detection model includes: Collect M sessions of multi-person Tai Chi exercise data; wherein the multi-person Tai Chi exercise data of each session includes: N global sample images synchronously photographed by the N cameras from the N different angles of the multi-person Tai Chi exercise in the target area and the three-dimensional coordinates of the UWB tags worn by the P participants in the session at each shooting time; M and P are preset values, and the participants in different sessions are the same, partially the same, or different; Preprocess each global sample image to obtain a preprocessed global sample image set; Calculate the two-dimensional projection coordinates of the UWB tags of each corresponding participant in each preprocessed global sample image according to the three-dimensional coordinates of each image and the UWB tags of each corresponding participant at the time of image capture in the preprocessed global sample image set; For each of the preprocessed global sample images, using OpenCV to obtain the rectangular frame position of the human body image of each corresponding participant; Selecting a sample image from the preprocessed global sample image set, and inputting the selected sample image and the two-dimensional projection coordinates of each corresponding participant in the image into the human figure detection model, and detecting the rectangular frame position of each corresponding participant's human body image in the sample image; Calculate a first loss function according to the rectangular frame position obtained by openCV and the detected rectangular frame position, and adjust the parameters of the human figure detection model; The steps of selecting sample images, inputting models, calculating the first loss function and adjusting parameters are repeated until the first loss function no longer decreases, thereby obtaining the trained human figure detection model.

[0012] Preferably, the training step of the key point detection model includes: Cropping each image in the preprocessed global sample image set according to the manually acquired rectangular frame position, and scaling the cropped rectangular image to the preset first size, thereby obtaining a local sample image set; For each local sample image in the local sample image set, use OpenCV to mark the key points of the human body and obtain the two-dimensional coordinates of the key points; Selecting a sample image from the local sample image set, and inputting the selected sample image into the key point detection model to detect the two-dimensional key point coordinates in the sample image; Calculate a second loss function according to the two-dimensional key point coordinates obtained by OpenCV and the detected two-dimensional key point coordinates, and adjust the parameters of the key point detection model; Repeat the steps of selecting sample images, inputting models, calculating the second loss function and adjusting parameters until the second loss function no longer decreases, thereby obtaining the trained key point detection model.

[0013] Preferably, the N scoring models correspond one-to-one to the N shooting angles; The N scoring models have the same structure, and the training steps include: According to the two-dimensional key point coordinates in the N local sample images corresponding to each participant in each session in the local sample image set, N key point coordinate time series of each participant in each session are generated, so as to obtain a time series sample set; the time series sample set includes MxP time series samples, and each of the time series samples includes N key point coordinate time series; According to the N local sample images corresponding to each participant in each session in the local sample image set, manually score the standardization of the participant's actions, thereby obtaining MxP scores; Selecting a time series sample from the time series sample set, sending N key point coordinate time series in the sample to N scoring models respectively, and outputting N predicted scores; According to the manual scores and the N predicted scores, N third loss functions are calculated respectively, and the corresponding N scoring models are adjusted respectively; Repeat the steps of selecting the time series samples, feeding the N scoring models, calculating the N third loss functions and adjusting the N scoring models until a preset number of iterations is reached to obtain the trained N scoring models.

[0014] Preferably, the key points of the human body of each participant include: head, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, the tip of the left middle finger, the tip of the right middle finger, left hip, right hip, left knee, right knee, left ankle, right ankle, left foot and right foot; The human figure detection model is a modified design of the YOLOv3 (You Only Look Once version 3, a target detection algorithm) network structure, guiding the model to focus on the two-dimensional projection coordinates of each participant; The key point detection model is OpenPose (a human posture estimation framework based on deep learning, which can estimate the key points of the human body in real time); The scoring model includes: an input layer, an LSTM layer, a fully connected layer and an output layer.

[0015] According to a second aspect of the present invention, a computer-readable storage medium is provided, storing a computer program that can be loaded by a processor and execute the method described above.

[0016] The present invention has the following beneficial effects: The present invention uses UWB positioning technology to track the three-dimensional coordinates of the UWB tags worn by each participant at each shooting time, which is helpful for quickly calculating the position of each participant in the image and convenient for distinguishing the identity of each participant in the image. The N key point coordinate time series of each participant are respectively input into N pre-trained scoring models, and the N scores obtained are averaged to obtain the final score of the participant. In this way, the participant can be observed and evaluated from N different angles, and the obtained score is more accurate. Therefore, the present invention solves the problem of scoring each participant using an image recognition method in a multi-person Tai Chi exercise. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the main steps of an embodiment of a method for identifying and scoring multi-person Tai Chi sports images of the present invention; Figure 2Schematic diagram of the main steps of training a human figure detection model in an embodiment of the present invention; Figure 3 Schematic diagram of the main steps of key point detection model training in an embodiment of the present invention; Figure 4 Schematic diagram of the main steps of training N scoring models in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0019] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] It should be noted that, in the description of the present invention, the terms "first" and "second" are only for the convenience of description, and do not indicate or imply the relative importance of the devices, elements or parameters, and therefore cannot be understood as limiting the present invention. In addition, the term "and / or" in the present invention is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally indicates that the associated objects before and after are in an "or" relationship.

[0021] Figure 1 Schematic diagram of the main steps of the embodiment of the method for identifying and scoring multi-person Tai Chi sports images of the present invention. Figure 1 The scoring method of this embodiment includes: Step S10: Use N cameras to synchronously capture N global images of multiple people doing Tai Chi in the target area from N different angles, and use UWB positioning technology to track the three-dimensional coordinates of the UWB tags worn by each participant at each shooting moment.

[0022] In this embodiment, N is a preset value; the target area is a rectangular sports field, and the outer frame is drawn in a striking color, and it can also be an elliptical or circular field according to actual conditions; N cameras are pre-fixed above or diagonally above the target area, and the position of each camera is higher than all participants to avoid overlapping portraits (maintaining an appropriate distance between people), and the shooting range of each camera is sufficient to cover the target area; each participant wears a UWB positioning tag on his chest, and each UWB positioning tag has an ID number as the ID of the corresponding participant; four UWB positioning base stations are set directly above the four vertices of the target area, and the positions are higher than all participants; the N cameras, all UWB positioning tags and four UWB positioning base stations are pre-calibrated at a unified time.

[0023] Step S20: According to the three-dimensional coordinates of the UWB tags worn by each participant at each shooting moment, N local images corresponding to each participant are cropped from the N global images.

[0024] This step specifically includes steps S21-S24: Step S21 , preprocessing N global images to obtain N preprocessed images.

[0025] Specifically, each image in the N channels of global images is denoised and scaled to a preset second size to obtain N channels of pre-processed images.

[0026] Step S22: Calculate the two-dimensional projection coordinates of the UWB tag of each participant in each image based on the N pre-processed images and the three-dimensional coordinates of the UWB tag of each participant at each shooting moment.

[0027] This step may specifically include steps S221-S223: Step S221: Calculate the vertical projection coordinates of the UWB tag of each participant in the plane where the target area is located at each shooting time according to the positions of the four UWB positioning base stations and the three-dimensional coordinates of the UWB tag of each participant at each shooting time.

[0028] In this step, the position of the UWB tag at each shooting moment is vertically projected onto the plane where the target area is located, and the position of each participant in the target area at each shooting moment is roughly determined.

[0029] Step S222: Use OpenCV to detect the coordinates of four vertices of the image corresponding to the target area in each image.

[0030] Step S223, according to the projection coordinates of each participant's UWB tag in the target area at each shooting moment, and the four vertex coordinates of the image corresponding to the target area in each image, calculate the two-dimensional projection coordinates of each participant's UWB tag in the image.

[0031] In this step, the approximate position of each participant is determined in the image, but each participant only corresponds to a point position, and the following step S23 is required to divide the complete image of each participant with a rectangular frame.

[0032] Step S23: According to the N pre-processed images and the two-dimensional projection coordinates of each participant, the position of the rectangular frame of each participant in each pre-processed image is detected using the pre-trained human figure detection model.

[0033] Step S24: crop the N pre-processed images according to the positions of the rectangular frames, and scale the cropped rectangular images to a preset first size to obtain N local images corresponding to each participant.

[0034] Step S30: According to the N local images corresponding to each participant, the coordinates of the human body key points of the participant in each local image are detected using a pre-trained key point detection model, thereby generating a time series of N key point coordinates of the participant.

[0035] In this embodiment, the key points of the human body of each participant include: head, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, fingertip of the left middle finger, fingertip of the right middle finger, left hip, right hip, left knee, right knee, left ankle, right ankle, left foot and right foot, a total of 17 key points.

[0036] One image is selected from the first path of partial images of a participant each time and input into the pre-trained key point detection model to obtain the two-dimensional coordinates of the human key points of the participant in each image of the first path of partial images (some undetectable ones can be filled with zeros). These key point coordinates are arranged in chronological order to generate the first key point coordinate time series (each value in the sequence is spliced ​​together from the coordinates of 17 points).

[0037] One image is selected from the second path of partial images of the participant each time and input into the key point detection model to obtain the two-dimensional coordinates of the human key points of the participant on each image in the second path of partial images, thereby generating the second key point coordinate time series of the participant.

[0038] The process is repeated until the Nth key point coordinate time series of the participant corresponding to the Nth local image is generated, and a total of N key point coordinate time series of the participant are obtained.

[0039] Step S40: Input the N key point coordinate time series of each participant into N pre-trained scoring models respectively to obtain N scores of the participant.

[0040] Step S50: average the N scores of each participant to obtain a final score of the participant.

[0041] In an optional embodiment, the human detection model is modified and designed based on the YOLOv3 network structure. The purpose of the modification is to fuse the image features with the two-dimensional projection coordinate features of each participant (UWB tag) and guide the model to focus on the two-dimensional projection coordinates of each participant so as to speed up the detection speed and distinguish the correspondence between the detected human body image and the participant ID.

[0042] Figure 2 Schematic diagram of the main steps of training a human detection model in an embodiment of the present invention. Figure 2 As shown, the main steps of human detection model training may include A10-A50: Step A10: Collect M sessions of multi-person Tai Chi exercise data.

[0043] Among them, the multi-person Tai Chi exercise data of each session includes: N global sample images synchronously shot by the N cameras from the N different angles of the multi-person Tai Chi exercise in the target area and the three-dimensional coordinates of the UWB tags worn by the P participants in the session at each shooting time; M and P are preset values, and the participants in different sessions can be the same, partially the same or different.

[0044] Step A20: preprocess each global sample image to obtain a preprocessed global sample image set.

[0045] Step A30, according to each image in the preprocessed global sample image set and the three-dimensional coordinates of the UWB tags of each corresponding participant at the time of image capture, calculate the two-dimensional projection coordinates of the UWB tags of each corresponding participant in each preprocessed global sample image.

[0046] Step A40: For each pre-processed global sample image, use OpenCV to obtain the rectangular frame position of the human body image of each corresponding participant.

[0047] Step A50, select a sample image from the preprocessed global sample image set, and input the selected sample image and the two-dimensional projection coordinates of each corresponding participant in the image into the human figure detection model to detect the rectangular frame position of each corresponding participant's human body image in the sample image.

[0048] Step A60: Calculate a first loss function according to the rectangular frame position obtained by openCV and the detected rectangular frame position, and adjust the parameters of the human figure detection model.

[0049] Step A70, go to step A50, until the first loss function no longer decreases, and obtain a trained human figure detection model.

[0050] In another optional embodiment, the key point detection model uses OpenPose, and its training method can use the pre-processed global sample image set obtained in the above steps A10-A20, and then perform the following Figure 3 Steps B10-B50 shown: Step B10: crop each image in the preprocessed global sample image set according to the manually acquired rectangular frame position, and scale the cropped rectangular image to a preset first size, thereby obtaining a local sample image set.

[0051] Step B20: For each local sample image in the local sample image set, use OpenCV to mark the key points of the human body and obtain the two-dimensional coordinates of the key points.

[0052] Step B30, selecting a sample image from the local sample image set, and inputting the selected sample image into the key point detection model to detect the two-dimensional key point coordinates in the sample image; Step B40, calculating a second loss function according to the two-dimensional key point coordinates obtained by OpenCV and the detected two-dimensional key point coordinates, and adjusting the parameters of the key point detection model; Step B50, go to step B30, until the second loss function no longer decreases, and obtain the trained key point detection model.

[0053] In another optional embodiment, the structures of the N scoring models are the same, and each scoring model may include: an input layer, an LSTM layer, a fully connected layer, and an output layer. The N scoring models correspond to the N shooting angles one by one. The training method can use the local sample image set generated in step B10 above, and then perform the following steps: Figure 4 Steps C10-C50 shown are used to train N scoring models together: Step C10, based on the two-dimensional key point coordinates in the N local sample images corresponding to each participant in each scene in the local sample image set, generate N key point coordinate time series for each participant in each scene, thereby obtaining a time series sample set; the time series sample set includes MxP time series samples, and each time series sample includes N key point coordinate time series.

[0054] Step C20: Based on the N local sample images corresponding to each participant in each session in the local sample image set, manually score the action standardization of the participant, thereby obtaining MxP scores.

[0055] Step C30: Select a time series sample from the time series sample set, send the N key point coordinate time series in the sample to N scoring models respectively, and output N predicted scores.

[0056] Step C40: Calculate N third loss functions respectively according to the manual ratings and the N predicted ratings, and adjust the corresponding N rating models respectively.

[0057] When calculating each third loss function, the manual rating is used as the true value, but the predicted value is the predicted rating output by the corresponding model.

[0058] Step C50, go to step C30, until the preset number of iterations is reached, and N trained scoring models are obtained.

[0059] Although each step in the above embodiments is described in the above-mentioned order, those skilled in the art can understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such an order, and they can be executed simultaneously (in parallel) or in a reverse order. These simple changes are within the scope of protection of the present invention.

[0060] Based on the above scoring method, the present invention further proposes an embodiment of a computer-readable storage medium. The storage medium of this embodiment stores a computer program that can be loaded by a processor and execute the method as described above.

[0061] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0062] Those skilled in the art should be able to appreciate that the method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0063] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A method for identifying and scoring multi-person Tai Chi sports images, characterized in that: The method comprises: N cameras are used to synchronously capture N global images of multiple people doing Tai Chi in the target area from N different angles, and UWB positioning technology is used to track the three-dimensional coordinates of the UWB tags worn by each participant at each shooting moment; According to the three-dimensional coordinates of the UWB tags worn by each participant at each shooting moment, N local images corresponding to each participant are cropped from the N global images; According to the N local images corresponding to each participant, the coordinates of the human body key points of the participant in each local image are detected using a pre-trained key point detection model, thereby generating a time series of N key point coordinates of the participant; Inputting the N key point coordinate time series of each participant into N pre-trained scoring models respectively to obtain N scores of the participant; averaging the N scores of each participant to obtain a final score of the participant; Wherein, N is a preset value.

2. The method for recognizing and scoring multi-person Tai Chi sports images according to claim 1, characterized in that: The target area is a rectangular sports field, and the outer frame is drawn in a striking color; The N cameras are pre-fixed and arranged above or diagonally above the target area, and the position of each camera is higher than all participants to avoid overlapping of portraits, and the shooting range of each camera is sufficient to cover the target area; Each participant wears a UWB positioning tag on his chest, and each UWB positioning tag has an ID number, which serves as the ID of the corresponding participant; Four UWB positioning base stations are set directly above the four vertices of the target area and are located higher than all participants; The N cameras, all UWB positioning tags and the four UWB positioning base stations are pre-calibrated at a unified time.

3. The method for recognizing and scoring multi-person Tai Chi sports images according to claim 2, characterized in that: The step of "cutting out N local images corresponding to each participant from the N global images according to the three-dimensional coordinates of the UWB tags worn by each participant at each shooting moment" includes: Preprocessing the N global images to obtain N preprocessed images; Calculate the two-dimensional projection coordinates of the UWB tag of each participant in each image based on the N pre-processed images and the three-dimensional coordinates of the UWB tag of each participant at each shooting time; According to the N pre-processed images and the two-dimensional projection coordinates of each participant, using a pre-trained human figure detection model, detecting the position of the rectangular frame of each participant in each pre-processed image; The N pre-processed images are cropped according to the positions of the rectangular frames, and the cropped rectangular images are scaled to a preset first size to obtain N local images corresponding to each participant.

4. The method for recognizing and scoring multi-person Tai Chi sports images according to claim 3, characterized in that: The step of "calculating the two-dimensional projection coordinates of the UWB tag of each participant in each image according to the N pre-processed images and the three-dimensional coordinates of the UWB tag of each participant at each shooting time" includes: Calculate the vertical projection coordinates of the UWB tag of each participant in the plane where the target area is located at each shooting time according to the positions of the four UWB positioning base stations and the three-dimensional coordinates of the UWB tag of each participant at each shooting time; OpenCV is used to detect the coordinates of the four vertices of the image corresponding to the target area in each image; According to the projection coordinates of each participant's UWB tag in the target area at each shooting moment and the four vertex coordinates of the image corresponding to the target area in each image, the two-dimensional projection coordinates of each participant's UWB tag in the image are calculated.

5. The method for recognizing and scoring multi-person Tai Chi sports images according to claim 1, characterized in that: The step of "detecting the coordinates of the human body key points of the participant in each local image using a pre-trained key point detection model according to the N local images corresponding to each participant, and then generating a time series of N key point coordinates of the participant" includes: Select one image from the first path of partial images of a participant at a time and input it into the pre-trained key point detection model to obtain the two-dimensional human body key point coordinates of the participant on each image in the first path of partial images, and then generate the first key point coordinate time series of the participant; Select one image from the second group of partial images of the participant at a time and input it into the key point detection model to obtain the two-dimensional coordinates of the human body key points of the participant on each image in the second group of partial images, thereby generating a second key point coordinate time series of the participant; The process is repeated until the Nth key point coordinate time series of the participant corresponding to the Nth local image is generated.

6. The method for recognizing and scoring multi-person Tai Chi sports images according to claim 3, characterized in that: The training steps of the human figure detection model include: Collect M sessions of multi-person Tai Chi exercise data; wherein the multi-person Tai Chi exercise data of each session includes: N global sample images synchronously photographed by the N cameras from the N different angles of the multi-person Tai Chi exercise in the target area and the three-dimensional coordinates of the UWB tags worn by the P participants in the session at each shooting time; M and P are preset values, and the participants in different sessions are the same, partially the same, or different; Preprocess each global sample image to obtain a preprocessed global sample image set; Calculate the two-dimensional projection coordinates of the UWB tags of each corresponding participant in each preprocessed global sample image according to the three-dimensional coordinates of each image and the UWB tags of each corresponding participant at the time of image capture in the preprocessed global sample image set; For each of the preprocessed global sample images, using OpenCV to obtain the rectangular frame position of the human body image of each corresponding participant; Selecting a sample image from the preprocessed global sample image set, and inputting the selected sample image and the two-dimensional projection coordinates of each corresponding participant in the image into the human figure detection model, and detecting the rectangular frame position of each corresponding participant's human body image in the sample image; Calculate a first loss function according to the rectangular frame position obtained by openCV and the detected rectangular frame position, and adjust the parameters of the human figure detection model; The steps of selecting sample images, inputting models, calculating the first loss function and adjusting parameters are repeated until the first loss function no longer decreases, thereby obtaining the trained human figure detection model.

7. The method for recognizing and scoring multi-person Tai Chi sports images according to claim 6, characterized in that: The training steps of the key point detection model include: Cropping each image in the preprocessed global sample image set according to the manually acquired rectangular frame position, and scaling the cropped rectangular image to the preset first size, thereby obtaining a local sample image set; For each local sample image in the local sample image set, use OpenCV to mark the key points of the human body and obtain the two-dimensional coordinates of the key points; Selecting a sample image from the local sample image set, and inputting the selected sample image into the key point detection model to detect the two-dimensional key point coordinates in the sample image; Calculate a second loss function according to the two-dimensional key point coordinates obtained by OpenCV and the detected two-dimensional key point coordinates, and adjust the parameters of the key point detection model; Repeat the steps of selecting sample images, inputting models, calculating the second loss function and adjusting parameters until the second loss function no longer decreases, thereby obtaining the trained key point detection model.

8. The method for recognizing and scoring multi-person Tai Chi sports images according to claim 7, characterized in that: The N scoring models correspond one-to-one to the N shooting angles; The N scoring models have the same structure, and the training steps include: According to the two-dimensional key point coordinates in the N local sample images corresponding to each participant in each session in the local sample image set, N key point coordinate time series of each participant in each session are generated, so as to obtain a time series sample set; the time series sample set includes MxP time series samples, and each of the time series samples includes N key point coordinate time series; According to the N local sample images corresponding to each participant in each session in the local sample image set, manually score the standardization of the participant's actions, thereby obtaining MxP scores; Selecting a time series sample from the time series sample set, sending N key point coordinate time series in the sample to N scoring models respectively, and outputting N predicted scores; According to the manual scores and the N predicted scores, N third loss functions are calculated respectively, and the corresponding N scoring models are adjusted respectively; Repeat the steps of selecting the time series samples, feeding the N scoring models, calculating the N third loss functions and adjusting the N scoring models until a preset number of iterations is reached to obtain the trained N scoring models.

9. The method for identifying and scoring multi-person Tai Chi sports images according to any one of claims 3-4 and 6-8, characterized in that: The key points of each participant’s body included: head, left shoulder, right shoulder, left elbow, right elbow, left wrist, right wrist, tip of the left middle finger, tip of the right middle finger, left hip, right hip, left knee, right knee, left ankle, right ankle, left foot, and right foot; The human figure detection model is a modified design of the YOLOv3 network structure, guiding the model to focus on the two-dimensional projection coordinates of each participant; The key point detection model is OpenPose; The scoring model includes: an input layer, an LSTM layer, a fully connected layer and an output layer.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 9.

Images