A method and device for monitoring the technical and tactical ability of a ball game and a storage medium

Abstract

The application relates to a kind of ball tactical ability monitoring method, device and storage medium.The method obtains original single-view video including multiple image frames;Based on each image frame, the camera internal and external orientation parameters are solved by optimizing using the correspondence between the 2D coordinates of the court key points preset by marking and the 3D coordinates of the same group of court key points obtained in advance;Different algorithms are used to obtain ball 2D sequence and human body key point 2D sequence respectively;Based on the optimized camera internal and external orientation parameters, ball 2D sequence and human body key point 2D sequence, the corresponding ball 3D position sequence and player 3D position sequence are obtained;Using the frame rate, ball 3D position sequence and player 3D position sequence obtained in advance, the tactical key data is calculated;Based on the original single-view video, the tactical key data is visually presented.Compared with the prior art, the application has the advantages of reducing the hardware environment building requirements and information processing load, improving the accuracy of player / ball 3D space-time feature analysis, etc.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, and in particular to a method, device, and storage medium for monitoring ball sports skills and tactics. Background Technology

[0002] Ball sports, as a globally popular genre, place high demands on athletes' technical and tactical abilities, as well as their ability to react quickly to changing situations. In high-level competitions, subtle differences in technique and tactics often determine the outcome. Therefore, accurate monitoring and analysis of athletes' technical and tactical abilities are crucial for improving training efficiency and competitive performance.

[0003] Traditional methods of ball sports training and match analysis rely on the observations of coaches and analysts. This approach is highly subjective, inefficient, and struggles to achieve large-scale and in-depth data analysis. With the development of computer vision technology, computers can accurately identify and track objects by analyzing images and videos, and this technology has been widely applied in various fields. In sports analysis, computer vision technology provides an efficient and objective means to support the monitoring and analysis of athlete performance. Specifically, key technologies for ball positioning and tracking and human posture detection and tracking based on computer vision make the monitoring and analysis of technical and tactical abilities possible. For example, Chinese patent CN116958872A discloses an intelligent auxiliary training method and system for badminton. This method first uses manually labeled RGB images of the court's feature points to map them to the real court positions in the world coordinate system, thereby calculating the camera's interior and exterior orientation parameters. For the badminton 2D positioning task, this invention uses mainstream deep learning algorithms to estimate the heatmap of the badminton's center position on the 2D image, obtaining the badminton's 2D pixel coordinates through a weighted summation. For the task of player 2D pose estimation, this invention adopts a top-down design, utilizing a target detection-target tracking-pose recognition approach. Specifically, firstly, a target detection algorithm is used to obtain bounding boxes for all identifiable human targets, and the court sidelines are used for filtering, retaining the bounding boxes corresponding to each player. Then, a 2D tracking algorithm is used to obtain the bounding box sequence for each player. The images within the bounding boxes are cropped and enlarged, and a 2D pose estimation algorithm is used to obtain the player's 2D pose. Subsequently, based on the 2D information of the shuttlecock and players in the monocular or binocular image coordinate system and camera parameters, triangulation is performed to obtain the 3D positions of the shuttlecock and players, and the trajectory of the shuttlecock and players is composed using continuous images or videos. However, existing methods still have the problems of high requirements for hardware environment setup and high information processing load. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology, such as high requirements for hardware environment setup and high information processing load, and to provide a method, device and storage medium for monitoring ball sports skills and tactics.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] According to a first aspect of the present invention, a method for monitoring ball game skills and tactics is provided, comprising the following steps: acquiring an original single-view video comprising multiple image frames; optimizing and solving the camera's interior and exterior orientation parameters based on each image frame by marking the correspondence between pre-defined 2D coordinates of key points on the court and pre-acquired 3D coordinates of the same group of key points on the court; acquiring ball game 2D sequences and human body key point 2D sequences respectively based on each image frame using different algorithms; acquiring corresponding ball game 3D position sequences and player 3D position sequences based on the optimized camera interior and exterior orientation parameters, the ball game 2D sequences, and the human body key point 2D sequences; calculating key skills and tactics data using the pre-acquired video frame rate, the ball game 3D position sequences, and the player 3D position sequences; and visually presenting the key skills and tactics data based on the original single-view video.

[0007] As a preferred technical solution, the process of optimizing the camera's interior and exterior orientation parameters specifically includes: selecting a first image frame, initializing and marking the 2D coordinates of the first key point of the first image frame, and calculating the corresponding camera interior and exterior orientation parameters using the 2D coordinates of the first key point and the corresponding 3D coordinates of the stadium key point; based on the 2D coordinates of the first key point, when the camera moves, using a deep learning-based tracking and matching algorithm to obtain the 2D coordinates of the second key points of the remaining image frames, and calculating the corresponding camera interior and exterior orientation parameters using the 2D coordinates of the second key points of each image frame and the corresponding 3D coordinates of the stadium key point.

[0008] As a preferred technical solution, the tracking and matching algorithm includes the LightGlue algorithm.

[0009] As a preferred technical solution, different algorithms are used to obtain 2D sequences of balls and 2D sequences of human key points. The specific process includes: using the TrackNetv3 algorithm to obtain 2D sequences of balls; using the RTMPose algorithm to obtain multiple 2D key point coordinate sequences of players; and using the OC-SORT algorithm to perform multi-target tracking on the 2D key coordinate sequence of each player to obtain the final 2D sequence of human key points.

[0010] As a preferred technical solution, the process of obtaining the corresponding 3D position sequence of the ball and the 3D position sequence of the player includes: selecting a key ball and marking the start image frame, end image frame, ball depth, and player depth of the key ball; calculating the corresponding 3D position sequence of the ball and the 3D position sequence of the player based on the ball depth, the player depth, the optimized camera interior and exterior orientation parameters, the 2D sequence of the ball, and the 2D sequence of the human body key points; and using kinematic theory to model and obtain the 3D trajectory of the key ball.

[0011] As a preferred technical solution, the key tactical data includes net crossing speed, attack angle, and defense distance.

[0012] As a preferred technical solution, the ball landing point is estimated based on the 3D trajectory of the key ball; the time and corresponding position of the ball trajectory on both sides of the net are obtained; the net crossing speed is calculated based on the time, the corresponding position and the video acquisition frame rate; the positions of the attacking player and the defending player in the starting image frame are obtained, and then a first line connecting the positions of the attacking player and the defending player, and a second line connecting the positions of the attacking player and the ball landing point are obtained; the attacking angle is calculated based on the first line and the second line; and the defending distance is calculated based on the position of the defending player and the ball landing point.

[0013] As a preferred technical solution, the key technical and tactical data is visualized, specifically by: adding special effects to the key technical and tactical data using Unity, adding them to the original single-view video using augmented reality methods, and displaying them in the corresponding positions through a display module.

[0014] According to a second aspect of the present invention, a ball game tactical ability monitoring device is provided. The device is used to implement the method described above. The device includes a 3D field calibration module, a ball and player positioning module, a tactical analysis module, and a 3D rendering module. The 3D field calibration module is used to acquire an original single-view video including multiple image frames, and based on each image frame, optimize the camera's interior and exterior orientation parameters by marking the correspondence between pre-defined 2D coordinates of key points on the field and pre-acquired 3D coordinates of the same group of key points on the field. The ball and player positioning module is used to acquire ball 2D sequences and human key point 2D sequences respectively based on each image frame using different algorithms, and acquire corresponding ball 3D position sequences and player 3D position sequences based on the optimized camera interior and exterior orientation parameters, the ball 2D sequences, and the human key point 2D sequences. The tactical analysis module is used to calculate key tactical data using the pre-acquired video frame rate, the ball 3D position sequences, and the player 3D position sequences. The 3D rendering module is used to visualize the key tactical data based on the original single-view video.

[0015] According to a third aspect of the present invention, a storage medium is provided having a program stored thereon, which, when executed, implements the method described thereon.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] 1. The ball game skills and tactics monitoring method provided by this invention only requires single-view video as input, without the need for wearable devices and special cameras or other hardware. It uses key points on the court combined with multiple algorithms to achieve high-precision 3D court calibration, obtain spatial information of the ball and players, and visualize key data of skills and tactics. It can effectively reduce the requirements for hardware environment construction and information processing load, while providing accurate and intuitive player skills and tactics, and facilitating the monitoring and analysis of player skills and tactics.

[0018] 2. This invention provides a single-view 3D court calibration process. Using the court markings as a reference, it optimizes the camera's interior and exterior orientation parameters by interactively annotating the 2D coordinates of key points and using a deep learning-based tracking and matching algorithm. This supports the calibration of dynamic cameras and effectively improves the accuracy of court calibration.

[0019] 3. This invention provides a single-view 3D positioning process for players and balls. Utilizing ball depth, player depth, optimized camera interior and exterior orientation parameters, ball 2D sequence, and human keypoint 2D sequence, it calculates the corresponding ball 3D position sequence and player 3D position sequence. That is, it uses pixel-precision position calculation and keypoint-granular player posture information, supplemented by an interactive depth annotation tool, to obtain the ball and player 3D position sequence from a single viewpoint. Based on kinematic modeling, it obtains multi-dimensional technical and tactical ability indicators such as net crossing speed, attack angle, and defensive distance, which are extensions of the kinematic information such as the basic ball and player movement trajectory, effectively improving the accuracy and comprehensiveness of player and ball 3D spatiotemporal feature analysis.

[0020] 4. Based on the acquisition of accurate data, this invention uses Unity to render realistic 3D animations of ball games and adds key event effects to provide users with an interface for analyzing and improving tactics. It can achieve accurate and multi-dimensional playback of ball game tactics indicators, and its visualization capabilities significantly improve the intuitiveness and convenience of tactical capability monitoring and analysis.

[0021] 5. The present invention can still achieve the beneficial effects of the provided technical solution under different ball trajectory modeling methods, different human skeleton key points, or different 3D rendering special effects methods. It can effectively improve the scene adaptability of ball sports tactical ability monitoring methods and devices, and is easy to promote. Attached Figure Description

[0022] Figure 1 A flowchart illustrating the method of this invention;

[0023] Figure 2 This is a schematic diagram of the execution flow of the device provided by the present invention. Detailed Implementation

[0024] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0025] Example 1

[0026] like Figure 1 As shown, this embodiment provides a method for monitoring ball game skills and tactics. This method, without the need for wearable devices or special cameras, can achieve high-precision 3D calibration of the court, accurate acquisition of ball trajectories and player postures, calculation and visualization of multi-dimensional key player skills and tactics data (i.e., multi-dimensional skills and tactics indicators), and 3D animation playback, among other functions. The method includes the following steps:

[0027] Step S1: Obtain the original single-view video including multiple image frames;

[0028] Step S2: Based on each image frame, the camera's interior and exterior orientation parameters are optimized by matching the pre-defined 2D coordinates of the key points on the court with the pre-acquired 3D coordinates of the same group of key points on the court.

[0029] Step S3: Based on each image frame, use different algorithms to obtain 2D sequences of spheres and 2D sequences of human key points respectively;

[0030] Step S4: Based on the optimized camera interior and exterior orientation parameters, the 2D sequence of the ball and the 2D sequence of human key points, obtain the corresponding 3D position sequence of the ball and the 3D position sequence of the player.

[0031] Step S5: Calculate key technical and tactical data by using the pre-acquired video to obtain the frame rate, ball 3D position sequence, and player 3D position sequence;

[0032] Step S6: Based on the original single-view video, visualize and present key technical and tactical data.

[0033] This embodiment also provides a ball game skills and tactics monitoring device for implementing the steps of the aforementioned method. Figure 2 The execution flow is illustrated. The device includes a 3D court calibration module, a ball and player positioning module, a tactical analysis module, and a 3D rendering module. Specifically, the 3D court calibration module implements steps S1-S2, the ball and player positioning module implements steps S3-S4, the tactical analysis module implements step S5, and the 3D rendering module implements step S6. The specific workflow of each module working together is as follows:

[0034] (1) Stadium 3D Calibration Module

[0035] The 3D stadium calibration module interactively marks key points on the stadium, using a deep learning-based tracking and matching algorithm to obtain the positions of key points in each image frame of the original video, and calculates the camera's interior and exterior orientation parameters to achieve stadium calibration. Specifically:

[0036] This module utilizes the Python OpenCV software library to implement an interactive graphical interface. Users sequentially mark the four corner points of the court sidelines and the two top corners of the net (if any), inputting the 2D coordinates of the key points in the image coordinate system and the corresponding 3D coordinates of the court key points into the court 3D calibration module. By matching the known 3D coordinates of the court key points in the world coordinate system with the 2D coordinates of the key points marked by the user in the image coordinate system, the camera's interior and exterior orientation parameters can be optimized.

[0037] When the camera moves, the module first initializes and marks the 2D coordinates of key points in the first image frame (e.g., selecting frame 0 as the first image frame) to obtain the 2D coordinates of the first key points. Then, it matches these coordinates with the corresponding 3D coordinates of the key points on the court and calculates the corresponding camera interior and exterior orientation parameters. Next, it uses the LightGlue algorithm to track and match the 2D positions of key points in the remaining image frames of the original video (i.e., obtains multiple 2D coordinates of second key points) and calculates the camera interior and exterior orientation parameters in the remaining image frames of the original video in the same way.

[0038] It is important to note that in this embodiment, key points refer to a series of points with specific physical meanings, such as the corner points of a sports field. Their 3D coordinates refer to their position in the world coordinate system, while their 2D coordinates refer to their position in the image coordinate system. Due to camera movement, the 2D coordinates of the key points in different frames will change; therefore, it is necessary to track and match the 2D coordinates of the key points in the remaining frames.

[0039] In practical applications, for image frame I i Given a video stream V = {I0, I1, ...}, the stadium 3D calibration module uses frame 0 to calibrate the camera's intrinsic and extrinsic parameters K and E, i.e., manually selecting the 2D coordinates of key points on the stadium. 2d , with known 3D coordinates of key points on the court 3d As input to the stadium 3D calibration module, it optimizes the camera's interior and exterior orientation parameters. If the camera moves, the LightGlue algorithm uses the 2D coordinates of key points from the previous time period. Calculate the 2D coordinates of the key points at the current moment. This allows us to obtain the 2D coordinates of key points on the court in each frame.

[0040] (2) Ball Games and Player Positioning Module

[0041] The ball and player localization modules utilize deep learning-based ball and player detection and tracking algorithms to obtain the 2D positions of the ball and players in each image frame of the original video. Subsequently, an interactive graphical interface is used to determine the start and end image frames of the keyball, and to label the ball's depth and player's depth, obtaining the 3D positions of the ball and players in the keyframes. Kinematic modeling is then used to obtain the 3D trajectory of the keyball. Specifically:

[0042] For 2D localization of balls: using image frames as input, the TrackNetv3 algorithm, finely tuned for different motion types, is used to obtain 2D sequences of balls;

[0043] For player 2D localization: using image frames as input, the RTMPose algorithm is used to perform 2D human pose estimation, that is, to obtain multiple player 2D key point coordinate sequences. Then, the OC-SORT algorithm is used for multi-target tracking to obtain the human key point 2D sequence.

[0044] This module utilizes the Python OpenCV software library to implement another interactive graphical interface for selecting key balls from image frames. Users mark the start and end image frames, ball depth, and player depth for the key ball. Using the camera's interior and exterior orientation parameters obtained from the court 3D calibration module, the 3D positions of the ball and player in each frame are calculated. Once the 2D coordinates, depth, and camera interior and exterior orientation parameters of the ball / player are determined, the 3D position can be calculated based on the 2D coordinates, depth, and camera parameters from multiple viewpoints. Considering the gravity, air resistance, and Magnus force generated by rotation experienced by the ball during its flight, the trajectory of the ball between the start and end image frames can be accurately modeled, and the landing point can be estimated.

[0045] In practical applications, the TrackNetv3 algorithm uses image frame I i As input, output 2D coordinates of the sphere. Each frame's 2D coordinates of the spheres form a 2D sequence of spheres x. ball The RTMPose algorithm uses image frames I i As input, the output is the coordinates of multiple player 2D keypoints in the scene. Next, the OC-SORT algorithm calculates the key point coordinate sequence x for each player. player Optimization is performed to obtain the final 2D sequence of human key points. The algorithms used in this module are characterized by high accuracy and low latency, and can be applied to efficient monitoring and analysis.

[0046] (3) Technical and tactical analysis module

[0047] The tactical analysis module calculates key tactical parameters based on the ball's 3D position sequence and the player's 3D position sequence. Both the ball's 3D position sequence and the player's 3D position sequence are time-series 3D data, containing the 3D positions of the ball and players at different times.

[0048] This module primarily calculates key tactical data based on the video frame rate and 3D position data obtained from the ball and player positioning modules. Key tactical data includes net-crossing speed, attack angle, and defensive distance. The calculation process for each key tactical data point includes:

[0049] First, based on the key ball 3D trajectory obtained from the ball and player positioning module, the ball's landing point is estimated;

[0050] Secondly, the time and corresponding position of the ball trajectory on both sides of the net are obtained; and the net crossing speed is calculated based on the time and corresponding position and the video frame rate. The net crossing speed is the average speed between two frames, that is, the distance of the ball multiplied by the frame rate.

[0051] Next, the positions of the attacking player and the defending player in the starting image frame are obtained, and then the first line connecting the positions of the attacking player and the defending player, and the second line connecting the position of the attacking player and the ball landing point are obtained. The attack angle is calculated from the angle between the two lines.

[0052] Finally, the defensive distance is calculated based on the distance between the defensive player's position and the ball's landing point in the initial image frame.

[0053] Specific examples are as follows:

[0054] The ball trajectory function f is a function of initial velocity v0, initial position x0, and time t, f(v0,x0,t). Given the initial position x0, the intermediate position x1, and the corresponding time t1, optimize the solution for the initial velocity v0:

[0055] v0= argmin(||x1 – f(v0,x0,t0)||) (1)

[0056] Where f is obtained by integrating at each time step:

[0057] f = ∫(v t +a t t)*tdt+x0 (2)

[0058] Wherein, acceleration a t Represented as:

[0059] a t = -g-λ d *norm(v)*v / m+λ l *norm(v)*cross(ω,v) / m (3)

[0060] In the formula, g is the acceleration due to gravity, and λ d The empirical resistance parameter is 0.00128625, λ l Let ω be the empirical parameter of the Magnus force, 0.0005145, ω be the horizontal vector, and m be the mass of the ball.

[0061] Therefore, the speed v of passing through the network net Represented as:

[0062] v net = ||x left - x right ||2*fps (4)

[0063] In the formula, x left x right The position of the balls at the moment they are positioned on either side of the net; fps is the frame rate.

[0064] The angle of attack is represented as:

[0065]

[0066] In the formula, x off For the attacking player position, x def For defensive player positions, The point where the ball lands is the location of the ball's landing point. The attack angle is the angle between the vectors connecting the attacking and defending players and the vectors connecting the defending player and the point where the ball lands.

[0067] The defensive distance d is represented as:

[0068]

[0069] In the formula, x def For defensive player positions, This indicates the point where the ball will land.

[0070] (4) 3D rendering module

[0071] The 3D rendering module utilizes Unity to design visualization effects for key tactical and technical data, overlaying them onto the original video in an augmented reality manner. Specifically:

[0072] This module uses Unity to add effects to key tactical data calculated in the tactical analysis module, and then adds them to the original video in an augmented reality manner. The module uses a Bézier curve to fit the start, middle, end, and expected landing point of the ball's 3D trajectory to control the trajectory of the arrow effects in a virtual 3D coordinate system. The player's 3D position at each moment and the expected landing point of the key ball are marked by a ring effect at the corresponding calculated position. Key tactical data such as net speed, attack angle, and defensive distance are presented in a Unity UI panel, with specific values ​​additionally displayed in the corresponding screen positions using TextMesh Pro.

[0073] In summary, the ball game skills and tactics monitoring method and device provided in this embodiment have the following advantages:

[0074] 1) Camera calibration is performed using key points on the court and supports dynamic camera calibration, resulting in better scene generalization capabilities;

[0075] 2) By using ball dynamics modeling and combining deep learning-based tracking and matching algorithms with interactive depth estimation algorithms, the 3D trajectory of the ball is obtained, which has higher accuracy for ball and player positioning;

[0076] 3) Using Unity to add special effects rendering to key technical and tactical data, it has complete functions and richer presentation forms of technical and tactical capabilities.

[0077] Example 2

[0078] This embodiment provides a storage medium on which a program is stored, which, when executed, implements the aforementioned method. The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine as a standalone software package, or entirely on a remote machine or server. In the context of this invention, a computer-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0079] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for monitoring ball game skills and tactics, characterized in that, Includes the following steps: Acquire raw single-view video including multiple image frames; Based on each image frame, the camera's interior and exterior orientation parameters are optimized by using the pre-defined 2D coordinates of key points on the court to correspond with the pre-acquired 3D coordinates of key points in the same group of courts. Based on each image frame, different algorithms were used to obtain 2D sequences of spheres and 2D sequences of human key points, respectively. Based on the optimized camera interior and exterior orientation parameters, the ball 2D sequence and the human body key point 2D sequence, the corresponding ball 3D position sequence and player 3D position sequence are obtained. Calculate key technical and tactical data by using the pre-acquired video frame rate, the ball's 3D position sequence, and the player's 3D position sequence; Based on the original single-view video, the key technical and tactical data are visualized. The process of optimizing the camera's interior and exterior orientation parameters specifically includes: Select the first image frame, initialize and mark the 2D coordinates of the first key point of the first image frame, and use the 2D coordinates of the first key point and the corresponding 3D coordinates of the ball field key point to calculate the corresponding camera interior and exterior orientation parameters. Based on the first key point 2D coordinates, when the camera moves, the second key point 2D coordinates of the remaining image frames are obtained using a deep learning-based tracking and matching algorithm. The corresponding camera interior and exterior orientation parameters are calculated using the second key point 2D coordinates of each image frame and the corresponding stadium key point 3D coordinates.

2. The method for monitoring ball game skills and tactics according to claim 1, characterized in that, The tracking and matching algorithm includes the LightGlue algorithm.

3. The method for monitoring ball game skills and tactics according to claim 1, characterized in that, Different algorithms were used to obtain 2D sequences of spheres and 2D sequences of human keypoints, respectively. The specific process included: Use the TrackNetv3 algorithm to obtain 2D sequences of spheres; The RTMPose algorithm is used to obtain the 2D key point coordinate sequences of multiple players, and the OC-SORT algorithm is used to perform multi-target tracking on the 2D key point coordinate sequence of each player to obtain the final 2D sequence of human key points.

4. The method for monitoring ball game skills and tactics according to claim 1, characterized in that, The process of obtaining the corresponding 3D position sequences of the ball and players includes: Select the key ball and mark its start image frame, end image frame, ball depth, and player depth; Based on the ball depth, the player depth, the optimized camera interior and exterior orientation parameters, the ball 2D sequence and the human body key point 2D sequence, calculate the corresponding ball 3D position sequence and player 3D position sequence; By using kinematic theory to model and obtain the 3D trajectory of the key ball.

5. The method for monitoring ball game skills and tactics according to claim 4, characterized in that, The key tactical data includes net crossing speed, attack angle, and defensive distance.

6. The method for monitoring ball game skills and tactics according to claim 5, characterized in that, The calculation process for the key tactical and technical data includes: Based on the 3D trajectory of the key ball, estimate the ball's landing point; Obtain the time and corresponding position when the ball's trajectory is on both sides of the net; The network speed is calculated based on the time, the corresponding location, and the video acquisition frame rate; The positions of the attacking player and the defending player in the starting image frame are obtained, and then a first line connecting the positions of the attacking player and the defending player, and a second line connecting the position of the attacking player and the point where the ball lands are obtained. Calculate the attack angle based on the first and second connecting lines; The defensive distance is calculated based on the defensive player's position and the ball's landing point.

7. The method for monitoring ball game skills and tactics according to claim 1, characterized in that, The key technical and tactical data are presented in a visual format, specifically including: Special effects are added to the key technical and tactical data using Unity, and then added to the original single-view video using augmented reality methods, which are then displayed in the corresponding positions through the display module.

8. A device for monitoring ball game skills and tactics, characterized in that, The device is used to implement the method as described in any one of claims 1-7, and the device includes a 3D court calibration module, a ball and player positioning module, a tactical analysis module, and a 3D rendering module, wherein: The stadium 3D calibration module is used to acquire the original single-view video including multiple image frames, and based on each image frame, it uses the pre-set 2D coordinates of the stadium key points to correspond with the pre-acquired 3D coordinates of the same group of stadium key points to optimize the solution of the camera's interior and exterior orientation parameters. The ball and player positioning module is used to obtain the ball 2D sequence and the human body key point 2D sequence respectively based on each image frame using different algorithms, and to obtain the corresponding ball 3D position sequence and player 3D position sequence based on the optimized camera interior and exterior orientation parameters, the ball 2D sequence and the human body key point 2D sequence. The technical and tactical analysis module is used to calculate key technical and tactical data by using the pre-acquired video frame rate, the ball 3D position sequence and the player 3D position sequence; The 3D rendering module is used to visualize the key technical and tactical data based on the original single-view video.

9. A storage medium having a program stored thereon, characterized in that, When the program is executed, it implements the method as described in any one of claims 1-7.

Images