Goal scoring recognition method and system based on stadium environment and player posture

Abstract

The present invention provides a goal scoring recognition method and system based on the stadium environment and the posture of the players. The method models the stadium environment information, utilizes a multi-target tracking algorithm to track the targets on the stadium, and simultaneously calls a posture estimation algorithm to perform skeleton key point detection on the tracked targets, thereby obtaining the continuous skeleton key point coordinates of each target on the stadium. After determining the shooting behavior, the real-time position is calculated, and the real-time position of the player in the bird's-eye view is calculated using affine transformation to achieve time-space positioning.

Description

Technical Field

[0001] The present disclosure relates to the field of computer vision technology, and in particular to a method and system for identifying a goal score based on a stadium environment and a player's posture. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] At present, with the development of deep learning technology, environmental modeling technology, such as SLAM (Simultaneous Localization and Mapping) has developed rapidly, and video processing and understanding technology has also become increasingly mature. In view of the fact that the scores in most ball games are determined by referees and staff on the scene, if it is completely dependent on manual scoring, it is subjective and easy to cause inaccurate scoring. The present invention considers applying video understanding technology to game scoring. For example, in basketball games, it is well known that, except for the free throw stage, the player's goal score is closely related to the position of his shot. At present, there is no automatic calculation method that combines stadium environment modeling with algorithms such as shooting target positioning to realize the player's goal score. In addition, the existing ball game scoring method through referee staff is overly dependent on on-site staff, manpower and subjectivity, which has a certain impact on the fairness of scoring in ball games. Summary of the invention

[0004] In order to solve the above problems, the present disclosure proposes a goal scoring recognition method based on the stadium environment and the posture of the players. The present disclosure combines the stadium environment modeling with the shooting target positioning algorithm to realize the automatic calculation of the players' goal scores, so as to improve the phenomenon that the scoring of the game is overly dependent on the on-site staff.

[0005] According to some embodiments, the present disclosure adopts the following technical solutions:

[0006] The goal scoring recognition method based on the stadium environment and the player's posture includes the following steps:

[0007] Capture panoramic video of the live game in the stadium in real time from a fixed perspective, and pre-process the acquired video data to convert it into a frame sequence;

[0008] Construct a two-dimensional coordinate system to track the target on the court and detect the target's skeleton key points;

[0009] Take the key frame coordinates to identify and locate the shooting behavior;

[0010] If it is determined that a shooting behavior occurs, the target information is recorded, and the current court image in the frame is transformed using affine transformation to obtain the position of the shooting behavior target in the transformed image;

[0011] The goal score is determined by using the position coordinates of the target where the shooting behavior occurs in the converted image.

[0012] According to other embodiments, the present disclosure also adopts the following technical solutions:

[0013] The goal scoring recognition system based on the stadium environment and personnel posture includes:

[0014] The video acquisition module includes a camera for real-time acquisition of panoramic video of the live game in the stadium at a fixed viewing angle;

[0015] The data processing module pre-processes the acquired video data and converts it into a frame sequence;

[0016] The behavior analysis, judgment and positioning module is used to construct a two-dimensional coordinate system, track the target on the court, detect the key points of the skeleton, and obtain the key frame coordinates to identify and locate the shooting behavior;

[0017] The scoring calculation module is used to transform the current court image in the frame using affine transformation, obtain the position of the shooting target in the transformed image, and use the position coordinates of the target where the shooting behavior occurs in the transformed image to determine the goal score.

[0018] Beneficial Effects

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The present disclosure considers combining the stadium environment modeling with the shooting target positioning algorithm to realize the automatic calculation of the player's goal score, uses the tracking and posture estimation algorithm to locate the target in time and space, calculates the goal score according to the relative relationship between the located target position and the stadium environment, and calculates the real-time position of the player who has shot in the figure according to the affine transformation calculation. The spatial and temporal positioning of the player's shooting behavior is realized. The goal score is calculated according to the relative position of the player and the three-point line of the stadium. The implementation of the algorithm can improve the problem that the current scoring of the game depends on manpower. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.

[0022] Figure 1 Schematic diagram of a goal scoring recognition method based on a stadium environment and a player's position and posture according to the present invention;

[0023] Figure 2 is a flow chart of a goal scoring recognition method based on a stadium environment and a player's position and posture disclosed in the present invention;

[0024] Figure 3 is a schematic diagram of a two-dimensional coordinate system in an original video frame of the present disclosure;

[0025] Figure 4 It is a schematic diagram before and after the affine transformation of the present invention. DETAILED DESCRIPTION

[0026] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0027] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0028] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0029] Example 1

[0030] The present disclosure provides a goal scoring recognition method based on the stadium environment and the posture of the players, which innovatively combines the stadium environment modeling with the current target tracking and posture estimation algorithms, locates the target position with shooting behavior through the algorithm, maps the target position and the stadium environment to a bird's-eye view through affine transformation to calculate the goal score.

[0031] This algorithm is based on two premises. First, it assumes that the video data to be identified is a panoramic video and there is no movement of the shooting angle. Second, it assumes that in the non-penalty stage, if the player shoots, a goal is assumed to be scored. The purpose of this algorithm is to calculate the specific score of the goal.

[0032] like Figure 1 as well as Figure 2 As shown, the goal scoring recognition method based on the stadium environment and the player's posture specifically includes the following steps:

[0033] Step S0: real-time acquisition of panoramic video of a live match in a stadium at a fixed viewing angle, and pre-processing of the acquired video data to convert it into a frame sequence;

[0034] Step S1: construct a two-dimensional coordinate system, track the target on the court and detect the key points of the skeleton;

[0035] Step S2: taking the key frame coordinates to identify and locate the shooting behavior. If it is determined that the shooting behavior occurs, the target information is recorded, and the current court image in the frame is transformed using affine transformation to obtain the position of the shooting behavior target in the transformed image;

[0036] Step S4: Determine the goal score using the position coordinates of the target where the shooting behavior occurs in the converted image.

[0037] In step S0, a panoramic video of a live match on the court is shot in real time using a camera at a fixed camera viewing angle, and the video data is saved.

[0038] The captured video data is obtained and preprocessed. For the game video data shot in the stadium, it is first converted into a frame sequence, recorded as {1,2,…,t-1,t,t+1,…,s}, with a total of s frames, where s is 500.

[0039] In step S1, when constructing a two-dimensional coordinate system, the short side of the video frame is used as the y-axis, the long side is used as the x-axis, and the positive direction is predefined. According to the two-dimensional coordinate system, Figure 3 As shown, the two-dimensional coordinates of any point in the frame can be obtained.

[0040] Next, we will track the target on the court and detect the key points of the skeleton. Specifically:

[0041] A multi-target tracking algorithm (taking the Towards Real-Time multi-target tracking algorithm as an example) is used to implement target tracking, and a posture estimation algorithm (taking OpenPose as an example) is used to perform skeleton key point detection on the tracked target.

[0042] Take the target tracking and skeleton key point detection of {t-1, t, …, s} frames as an example, the details are as follows:

[0043] (1) Assuming that the detection algorithm is called once every t-1 frame during the tracking process to correct the true position of each target detection frame, the Yolo-v3 detection algorithm is called in the t-1 frame to obtain the detection frames of several targets;

[0044] (2) Use the Towards Real-Time multi-target tracking algorithm to predict the detection box of each target in the next frame, i.e., the tth frame;

[0045] (3) Using feature matching technology, the target features in each predicted detection frame in the t-th frame are matched with the target features in each detection frame in the t-1-th frame. The detection frame of each target in the t-th frame is determined according to the matching similarity. The determination of the detection frame of each target in the subsequent frames is deduced in this way until the next detection algorithm is called.

[0046] (4) After the target detection frame of each frame is determined, the OpenPose algorithm is called to estimate the pose of the target in the detection frame to obtain the coordinate data of the skeleton key points of each target;

[0047] (5) The above operations can be executed cyclically to track each target in the video and detect skeleton key points.

[0048] The target detection frame in each frame obtained by the target tracking algorithm is recorded as in, represents the detection box of target j in the i-th frame, The width, height and upper left corner coordinate information of the detection box are stored in s represents the total number of frames in the video, and n represents the total number of targets in the video. The coordinate information of the skeleton key points obtained by calling the pose estimation algorithm for the target in the detection frame of target j in the i-th frame is recorded as Among them, m represents the total number of key points detected by the posture estimation algorithm, and o has a corresponding relationship with the key points of various parts of the human body.

[0049] In step S3, taking target j as an example, the current frame and the previous n-1 consecutive frames are used as key frames in the scheme, and the specific method for determining whether there is a shooting behavior is as follows:

[0050] Taking target j as an example, take the key frame, that is, take the coordinates of the key points of the elbow and shoulder joints of the current frame and the previous n-1 consecutive frames. If the average height of the elbow joint of target j for n consecutive frames is higher than the average height of the shoulder joint, it is determined that target j has a shooting behavior in the current frame.

[0051] Taking target j as an example, take the key frame, that is, take the coordinates of the key points of the elbow and shoulder joints of the current frame and the previous n-1 consecutive frames. If the average height of the elbow joint of target j for n consecutive frames is lower than the average height of the shoulder joint, it is determined that target j has no shooting behavior in the current frame.

[0052] When it is detected that a target has shot in a certain frame, the current detection frame information of the target is recorded, and the court image in the frame is converted into a bird's-eye view using affine transformation, and the two-dimensional coordinate system of the bird's-eye view is obtained.

[0053] That is, when it is detected in step S3 that the target j in the i-th frame has a shooting behavior, the detection frame information of the target at this moment is recorded, and the court captured in the frame is converted into a bird's-eye view using affine transformation.

[0054] As shown in Figure 4 the left figure in Figure 4 is the court diagram and its two-dimensional coordinate system from a certain perspective captured in the video, Figure 4 and the right figure in

[0055]

[0056] is the bird's-eye view of the court and its two-dimensional coordinate system after affine transformation. The affine transformation formula is as follows:

[0057] Among them, x′ and y′ represent the coordinates of the points in the original video frame represented by the coordinate system established in step S1, and x and y represent the coordinates of the corresponding points after affine transformation. Furthermore, using the affine transformation formula, the position of the target j with a shooting behavior in the bird's-eye view can be obtained. Given that its detection box in the original video frame is

[0058] Specify its position as (x + w / 2, y + h), and then substitute it into the affine transformation formula to obtain the position of target j in the bird's-eye view.

[0059] (1) First, according to the coordinate system in the bird's-eye view obtained by affine transformation, as shown in Figure 4 the right figure in

[0060] (2) As shown in Figure 4 the right figure in

[0061] (3) Similarly, if the player is in the right half court, draw a line parallel to the x-axis through point D. This line intersects the three-point line of this half court at point P(x4, y4). As shown in the figure, if x4 > x1, it is a three-point shot; otherwise, it is a two-point shot.

[0062] Embodiment 2

[0063] The present disclosure provides a goal scoring recognition system based on the court environment and the pose of personnel, specifically including:

[0064] The video acquisition module includes a camera for real-time acquisition of panoramic video of the live game in the stadium at a fixed viewing angle;

[0065] The data processing module pre-processes the acquired video data and converts it into a frame sequence;

[0066] The behavior analysis, judgment and positioning module is used to construct a two-dimensional coordinate system, track the target on the court, detect the key points of the skeleton, and obtain the key frame coordinates to identify and locate the shooting behavior;

[0067] The scoring calculation module is used to transform the current court image in the frame using affine transformation, obtain the position of the shooting target in the transformed image, and use the position coordinates of the target where the shooting behavior occurs in the transformed image to determine the goal score.

[0068] Since the present disclosure is based on two premises, one is that it is assumed that the video data to be identified is a panoramic video and there is no movement of the shooting angle, and the other is that it is assumed that a goal is scored if a player shoots during the non-free throw phase, a camera is used to capture a panoramic video of the live game on the court in real time at a fixed angle.

[0069] The above-mentioned goal scoring recognition system based on the stadium environment and the player's posture performs the following method steps:

[0070] Video data preprocessing: For the game video data shot in the stadium, first convert it into a frame sequence, recorded as {1,2,…,t-1,t,t+1,…,500}, a total of 500 frames;

[0071] Step S1: Construct a two-dimensional coordinate system. Figure 3 As shown, the short side of the video frame is used as the y-axis, the long side is used as the x-axis, and the predefined positive direction is also as Figure 3 As shown, according to the two-dimensional coordinate system, the two-dimensional coordinates of any point in the frame can be obtained;

[0072] Step S2: Track the target on the court and detect the skeleton key points. Use the Towards Real-Time multi-target tracking algorithm to achieve target tracking, and use the OpenPose pose estimation algorithm to detect the skeleton key points of the tracked target. Take the target tracking and skeleton key point detection of {1,2,…,t-1,t,t+1,…,500} frames as an example, as follows:

[0073] (1) Assuming that the detection algorithm is called once every 10 frames during the tracking process to correct the true position of each target detection frame, the Yolo-v3 detection algorithm is called in the first frame to obtain the detection frames of several targets;

[0074] (2) Use the Towards Real-Time multi-target tracking algorithm to predict the detection box of each target in the second frame;

[0075] (3) Using feature matching technology, the target features in each predicted detection frame in the second frame are matched with the target features in each detection frame in the first frame, and the detection frame of each target in the second frame is determined according to the matching similarity. The determination of the detection frame of each target in the subsequent frames is deduced in this way until the next detection algorithm is called;

[0076] (4) After the target detection frame of each frame is determined, the OpenPose algorithm is called to estimate the pose of the target in the detection frame to obtain the coordinate data of the skeleton key points of each target;

[0077] (5) The above operations can be executed cyclically to track each target in the video and detect skeleton key points.

[0078] The target detection frame in each frame obtained by the target tracking algorithm is recorded as in, represents the detection box of target j in the i-th frame, The width, height and upper left corner coordinate information of the detection box are stored in 500 represents the total number of frames in the video, and 20 represents the total number of targets in the video. The coordinate information of the skeleton key points obtained by calling the pose estimation algorithm for the target in the detection frame of target j in the i-th frame is recorded as Among them, m=18 represents the total number of key points detected by the posture estimation algorithm, o corresponds to the key points of each part of the human body, and 1 to 18 correspond to nose, neck, right shoulder, right elbow, right wrist, left shoulder, left elbow, left wrist, right hip, right knee, right ankle, left hip, left knee, left ankle, right eye, left eye, right ear, and left ear respectively.

[0079] Step S3: Shooting behavior determination. Use the target's skeleton key point coordinates to determine whether a shooting behavior occurs. Take target j as an example, take the elbow and shoulder key point coordinates of the current frame and the previous 5 consecutive frames. If the average elbow height of target j for 6 consecutive frames is higher than the average shoulder height, it is determined that target j has a shooting behavior in the current frame. Otherwise, target j does not have a shooting behavior in the current frame.

[0080] Step S4: Affine transformation. When it is detected in step S3 that target j has a shooting behavior in the i-th frame, the detection frame information of the target at this moment is recorded. Affine transformation is used to convert the stadium captured in the frame into a bird's-eye view, such as Figure 4As shown in the figure, on the left is the court diagram and its two-dimensional coordinate system from a certain perspective captured in the video, and on the right is the aerial view of the court and its two-dimensional coordinate system after affine transformation. The affine transformation formula is as follows. Here, x′ and y′ represent the coordinates of the points in the original video frame expressed in the coordinate system established in step S1, and x and y represent the coordinates of the corresponding points after affine transformation.

[0081]

[0082] Furthermore, using the affine transformation formula, the position of the target j with a shooting behavior in the aerial view can be obtained. Given that its detection box in the original video frame is Specify its position as (x + w / 2, y + h), and then substitute it into the affine transformation formula to obtain the position of target j in the aerial view.

[0083] Step S5: Goal scoring determination. Using step S4, the position coordinates of the target with a shooting behavior in the aerial view can be obtained, denoted as D1(x1, y1). This coordinate is expressed in the coordinate system established in the aerial view. Determine the score based on the distances between D1(x1, y1) and the half-court dividing line and the three-point line. The steps are as follows:

[0084] (1) First, according to the coordinate system in the aerial view obtained by affine transformation, as shown in the right figure of Figure 4 Denote the equation of the half-court dividing line after affine transformation as x = x2. Judge the size relationship between the x-axis coordinate of D1 and the x-axis coordinate of the half-court dividing line. If x1 < x2, the player is in the left half-court; otherwise, the player is in the right half-court.

[0085] (2) As shown in the right figure of Figure 4 If the player is in the left half-court, draw a line parallel to the x-axis through point D. This line intersects the three-point line of this half-court at point Q(x3, y3), as shown in the figure. If x3 < x1, it is a three-pointer; otherwise, it is a two-pointer.

[0086] (3) Similarly, if the player is in the right half-court, draw a line parallel to the x-axis through point D. This line intersects the three-point line of this half-court at point P(x4, y4), as shown in the figure. If x4 > x1, it is a three-pointer; otherwise, it is a two-pointer.

[0087] The present disclosure proposes an algorithm for modeling the stadium environment information, using the tracking and posture estimation algorithm to perform the target spatiotemporal positioning, and calculating the goal score according to the relative relationship between the positioned target position and the stadium environment. First, the video to be identified is preprocessed to obtain a frame sequence, and a two-dimensional coordinate system is constructed on the original frame to locate the position of the players on the court. Then, the multi-target tracking algorithm is used to track the targets on the court, and the posture estimation algorithm is called to perform skeleton key point detection on the tracked targets to obtain the continuous skeleton key point coordinates of each target on the court. Next, the skeleton key point coordinates of the target in consecutive n frames are used to determine whether a shooting behavior occurs. If a shooting behavior occurs, the target's detection frame information is used to calculate its real-time position, and then the affine transformation is used to convert the stadium map in the original video frame into a bird's-eye view. Similarly, the real-time position of the player in the bird's-eye view can be obtained according to the affine transformation calculation. So far, the above operation realizes the spatiotemporal positioning of the player's shooting behavior. Finally, the goal score of the target's shooting behavior can be calculated according to the relative position relationship between the position of the target in the bird's-eye view and the half-court line and the three-point line. We use the system for actual testing. The above method has the characteristics of being general and practical. The algorithm innovatively combines stadium environment modeling with current target tracking and posture estimation algorithms. It locates the target position with shooting behavior through the algorithm, maps the target position and the stadium environment to a bird's-eye view through affine transformation to calculate the goal score.

[0088] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0089] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0090] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0092] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

[0093] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. Goal scoring recognition method based on the stadium environment and player posture, It is characterized in that The following steps are involved: Capture panoramic video of the live game in the stadium in real time from a fixed perspective, and pre-process the acquired video data to convert it into a frame sequence; Construct a two-dimensional coordinate system to track targets on the court and detect key points of the skeleton; The key frame coordinates are used to identify and locate the shooting behavior. If it is determined that a shooting behavior occurs, the target information is recorded, and the current court image in the frame is transformed using affine transformation to obtain the position of the shooting behavior target in the transformed image. The goal score is determined by using the position coordinates of the target where the shooting behavior occurs in the converted image; The construction of two-dimensional coordinates is specifically as follows: Preprocess the video data, convert the obtained game video data into a frame sequence, recorded as {1,2,…,t-1,t,t+1,…,s}, with a total of s frames; use the short side of the video frame as the y-axis, the long side as the x-axis, predefine the positive direction, construct a two-dimensional coordinate, and obtain the two-dimensional coordinate of any point in the frame according to the two-dimensional coordinate system; The multi-target tracking algorithm is used to track the target, and the posture estimation algorithm is used to detect the skeleton key points of the tracked target.

2. The goal scoring recognition method based on the stadium environment and the player's posture as claimed in claim 1, It is characterized in that The target detection frame in each frame obtained by the target tracking algorithm is recorded as ,in, represents the detection box of target j in the i-th frame, The width, height and upper left corner coordinate information of the detection box are stored in , s represents the total number of frames in the video, and n represents the total number of objects in the video.

3. The goal scoring recognition method based on the stadium environment and the player's posture as claimed in claim 1, It is characterized in that The specific method for judging shooting behavior is: The target's skeleton key point coordinates are used to determine whether a shooting behavior occurs. A target is selected, and the coordinates of the elbow and shoulder joint key points of the current frame and the previous n-1 consecutive frames are taken. If the average elbow joint height of the target for n consecutive frames is higher than the average shoulder joint height, it is determined that the target has a shooting behavior in the current frame.

4. The goal scoring recognition method based on the stadium environment and the player's posture as claimed in claim 3, It is characterized in that The coordinates of the key points of the elbow and shoulder joints of the current frame and the previous n-1 consecutive frames are obtained. If the average height of the elbow joint of the target in n consecutive frames is lower than the average height of the shoulder joint, it is determined that the target has no shooting behavior in the current frame.

5. The goal scoring recognition method based on the stadium environment and the player's posture as claimed in claim 1, It is characterized in that When it is detected that a target has shot in a certain frame, the current detection frame information of the target is recorded, and the court image in the frame is converted into a bird's-eye view using affine transformation, and the two-dimensional coordinate system of the bird's-eye view is obtained.

6. The goal scoring recognition method based on the stadium environment and the player's posture as claimed in claim 5, It is characterized in that The affine transformation formula is as follows: = in, , Represents the coordinates of the midpoint of the original video frame represented by a two-dimensional coordinate system constructed based on the video data frame. , Represents the coordinates of corresponding points after affine transformation.

7. The method for identifying a goal based on the stadium environment and the position of a player according to claim 6, It is characterized in that The affine transformation formula is used to obtain the position of the target with shooting behavior in the bird's-eye view image. Specifically, it is known that the detection frame of the target in the original video frame is , specifying its location as , and then substitute the affine transformation formula to obtain the position coordinates of the target in the bird's-eye view.

8. The goal scoring recognition method based on the stadium environment and the player's posture as claimed in claim 1, It is characterized in that The position coordinates of the target that performs the shooting action in the bird's-eye view are used to represent the position coordinates in a two-dimensional coordinate system constructed in the bird's-eye view, and the goal score is determined based on the distances between the position coordinates and the half-court dividing line and the three-point line.

9. Goal scoring recognition system based on the stadium environment and player posture, It is characterized in that include: The video acquisition module includes a camera for real-time acquisition of panoramic video of the live game in the stadium at a fixed viewing angle; The data processing module pre-processes the acquired video data and converts it into a frame sequence; The behavior analysis, judgment and positioning module is used to construct a two-dimensional coordinate system, track the target on the court, detect the key points of the skeleton, and obtain the key frame coordinates to identify and locate the shooting behavior; A scoring calculation module is used to transform the current court image in the frame by using affine transformation, obtain the position of the target of the shooting behavior in the transformed image, and determine the goal score by using the position coordinates of the target of the shooting behavior in the transformed image; The construction of two-dimensional coordinates is specifically as follows: Preprocess the video data, convert the obtained game video data into a frame sequence, recorded as {1,2,…,t-1,t,t+1,…,s}, with a total of s frames; use the short side of the video frame as the y-axis, the long side as the x-axis, predefine the positive direction, construct a two-dimensional coordinate, and obtain the two-dimensional coordinate of any point in the frame according to the two-dimensional coordinate system; The multi-target tracking algorithm is used to track the target, and the posture estimation algorithm is used to detect the skeleton key points of the tracked target.

Images