A computer vision-based running violation detection system

Abstract

The application discloses a running rule violation detection system based on computer vision and relates to the technical field of running rule violation detection, solves the technical problems that it is difficult for a referee to accurately determine whether a player has a rule violation in starting and that a pressure sensor is inaccurate in detecting a sprint, the video data is collected by a data collection module, the video data is sent to a data processing module, the data processing module receives the video data, obtains a rule violation label according to the video data and a rule violation detection model, and the rule violation label is identified, when a player has a rule violation, the player's result is cancelled, whether a player has a rule violation in starting is determined by using computer vision, and the accuracy of detecting a sprint is improved.

Description

Technical Field

[0001] This invention belongs to the field of computer vision and relates to running violation detection technology, specifically a running violation detection system based on computer vision. Background Technology

[0002] The purpose of a start is to generate tremendous forward momentum, quickly get the body out of a stationary state, and create favorable conditions for acceleration after the start. Sprints involve relatively short distances, and in primary and secondary school physical education teaching and standardized tests, the 50-meter or 100-meter race is commonly used to assess students' speed and sprinting ability. Therefore, the quality of the starting motion is a crucial factor directly affecting sprint performance.

[0003] Currently, starting violations occur frequently in sprint races. It's difficult for referees to accurately determine whether an athlete has violated the rules using only the naked eye, causing them considerable inconvenience. At the start, the starting blocks are equipped with pressure sensors. If these sensors detect the set pressure limit within 0.1 seconds, it's considered a false start. However, research has found that detecting this based on toe movements is not always accurate.

[0004] To address this, a computer vision-based running violation detection system is proposed. Summary of the Invention

[0005] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, this invention proposes a computer vision-based running violation detection system, which solves the problems that referees find it difficult to accurately determine whether athletes have violated starting rules by relying on the naked eye, and the inaccuracy of using pressure sensors to detect false starts.

[0006] To achieve the above objectives, an embodiment of the first aspect of the present invention provides a computer vision-based running violation detection system, comprising a data acquisition module, a data processing module, and a data processing module; the modules interact with each other based on digital signals.

[0007] The data acquisition module is used to acquire video data; wherein, the video data includes frontal video data, overhead video data, and side video data;

[0008] And send the video data to the data processing module;

[0009] The data processing module is used to receive the video data and obtain violation tags based on the video data and the violation detection model.

[0010] The system also identifies the violation tags and disqualifies athletes from performing violations; the violation detection model is based on an artificial intelligence model.

[0011] Preferably, the data acquisition module includes three video acquisition devices;

[0012] The video capture devices are fixedly installed directly in front of, above, and to the side of the athlete's starting position.

[0013] Preferably, the data acquisition module acquires video data, and the specific process includes:

[0014] The data acquisition module obtains the time of the gunshot and uses the time of the gunshot as the reference time.

[0015] The data acquisition module acquires video data taken by the video acquisition device at the front position 0.1 seconds after the reference time, and marks it as the front video data;

[0016] The data acquisition module acquires video data captured by the video acquisition device directly above at a reference time 0.1 seconds later, and marks it as the video data above.

[0017] The data acquisition module acquires video data captured by the video acquisition device at the front and side positions 0.1 seconds after the reference time, and marks it as side video data;

[0018] The data acquisition module sends the front video data, the top video data, and the side video data to the data processing module.

[0019] Preferably, the violation label is obtained based on the video data and the violation detection model. The specific process includes:

[0020] The data processing module receives the video data;

[0021] Set the position of the reference object in the foreground video data as the origin, draw a straight line parallel to the horizontal plane through the origin, and this straight line is the x-axis. The straight line perpendicular to the x-axis through the origin is the z-axis, and establish the first rectangular coordinate system xoz.

[0022] Set the position of the reference object in the video data above as the origin, establish a second rectangular coordinate system xoy, set the x-axis in the first rectangular coordinate system as the x-axis of the second rectangular coordinate system, and set the line passing through the origin and perpendicular to the x-axis as the y-axis.

[0023] Set the position of the reference object in the side video data as the origin, establish a third rectangular coordinate system zoy, set the z-axis in the first rectangular coordinate system as the z-axis in the third rectangular coordinate system, and set the y-axis in the second rectangular coordinate system as the y-axis in the third rectangular coordinate system.

[0024] Set the position of the reference object as the origin, and establish a spatial rectangular coordinate system based on the first rectangular coordinate system, the second rectangular coordinate system, and the third rectangular coordinate system;

[0025] A three-dimensional model of the athlete is established based on the video data and the spatial rectangular coordinate system;

[0026] The athlete's head, left elbow, right elbow, left knee, and right knee are labeled in the 3D model;

[0027] N frames are obtained from the video data on average; where N is an integer greater than or equal to 2; where the image at the reference time is the first frame and the image at the last time is the last frame.

[0028] Each frame of the image is numbered according to the time sequence, and the number is represented by n; where n takes the value 1, 2...N;

[0029] The position coordinates of each frame of the image are obtained according to the spatial rectangular coordinate system; wherein, the position coordinates include the head coordinates, left elbow coordinates, right elbow coordinates, left knee coordinates, and right knee coordinates;

[0030] The head coordinates are marked as A. n (x n ,y n ,z n )

[0031] The coordinate of the left elbow is marked as B. 左n (x 左n ,y 左n ,z 左n )

[0032] The coordinate of the right elbow is marked as B. 右n (x 右n ,y 右n ,z 右n )

[0033] The coordinates of the left knee are marked as C. 左n (x 左n ,y 左n ,z 左n )

[0034] The coordinates of the right knee are marked as C. 右n (x 右n ,y 右n ,z 右n )

[0035] The position coordinates of each frame are combined into a coordinate group, resulting in a total of N coordinate groups.

[0036] Integrate the N sets of coordinates into the original data;

[0037] Obtain the violation detection model from the data processing module;

[0038] The raw data is input into the violation detection model to obtain the corresponding violation label;

[0039] The data processing module identifies the illegal labels;

[0040] When the violation label identifies an athlete as having committed a violation, the corresponding athlete's score will be cancelled.

[0041] Preferably, the value of the violation label is 0 or 1. When the violation label is 0, it means that the corresponding athlete has committed a violation. When the violation label is 1, it means that the corresponding athlete has not committed a violation.

[0042] Preferably, the violation detection model is established based on an artificial intelligence model, and the specific process includes:

[0043] Obtain standard training data from the data processing module;

[0044] The artificial intelligence model is trained using standard training data, and the trained artificial intelligence model is then marked as a violation detection model.

[0045] Artificial intelligence models include deep convolutional neural network models or RBF neural network models, which have strong nonlinear fitting capabilities.

[0046] Preferably, the data acquisition module is communicatively and / or electrically connected to the data processing module.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This invention collects video data through a data acquisition module and sends the video data to a data processing module. The data processing module receives the video data, obtains violation tags based on the video data and a violation detection model, and identifies the violation tags. When an athlete commits a violation, the athlete's score is cancelled. This invention enables the use of computer vision to determine whether an athlete has committed a false start violation, thus improving the accuracy of false start detection. Attached Figure Description

[0049] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0050] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] like Figure 1 As shown, a computer vision-based running violation detection system includes a data acquisition module, a data processing module, and a data processing module; the modules interact with each other using digital signals.

[0052] The data acquisition module is used to acquire video data; wherein, the video data includes frontal video data, overhead video data, and side video data;

[0053] And send the video data to the data processing module;

[0054] The data processing module is used to receive the video data and obtain violation tags based on the video data and the violation detection model.

[0055] The system also identifies the violation tags and disqualifies athletes from performing violations; the violation detection model is based on an artificial intelligence model.

[0056] In this embodiment, the data acquisition module includes three video acquisition devices;

[0057] The video capture devices are fixedly installed directly in front of, above, and to the side of the athlete's starting position.

[0058] The data acquisition module acquires video data, specifically through the following process:

[0059] The data acquisition module obtains the time of the gunshot;

[0060] The moment the gun fired is used as the baseline time;

[0061] For example:

[0062] The gunshot was fired at 9:10:00, and the reference time is 9:10:00.

[0063] The data acquisition module acquires video data taken by the video acquisition device at the front position 0.1 seconds after the reference time, and marks it as the front video data;

[0064] The data acquisition module acquires video data captured by the video acquisition device directly above at a reference time 0.1 seconds later, and marks it as the video data above.

[0065] The data acquisition module acquires video data captured by the video acquisition device at the front and side positions 0.1 seconds after the reference time, and marks it as side video data;

[0066] It needs to be further explained that the human body has a reaction time. When you hear a command or signal, it is transferred to the central nervous system and then controls some muscles to start moving. There is definitely a time difference between these processes. The reaction time of an average person is more than 0.2 seconds, but well-trained athletes and soldiers can shorten it, but it will not be less than 0.1 seconds. This is determined by the principle of the human nervous system and cannot be reduced further.

[0067] The data acquisition module sends the front video data, the top video data, and the side video data to the data processing module.

[0068] The process of obtaining violation tags based on the video data and violation detection model includes:

[0069] The data processing module receives the video data;

[0070] Set the position of the reference object in the foreground video data as the origin, draw a straight line parallel to the horizontal plane through the origin, and this straight line is the x-axis. The straight line perpendicular to the x-axis through the origin is the z-axis, and establish the first rectangular coordinate system xoz.

[0071] Set the position of the reference object in the video data above as the origin, establish a second rectangular coordinate system xoy, set the x-axis in the first rectangular coordinate system as the x-axis of the second rectangular coordinate system, and set the line passing through the origin and perpendicular to the x-axis as the y-axis.

[0072] Set the position of the reference object in the side video data as the origin, establish a third rectangular coordinate system zoy, set the z-axis in the first rectangular coordinate system as the z-axis in the third rectangular coordinate system, and set the y-axis in the second rectangular coordinate system as the y-axis in the third rectangular coordinate system.

[0073] Set the position of the reference object as the origin, and establish a spatial rectangular coordinate system based on the first rectangular coordinate system, the second rectangular coordinate system, and the third rectangular coordinate system;

[0074] A three-dimensional model of the athlete is established based on the video data and the spatial rectangular coordinate system;

[0075] The athlete's head, left elbow, right elbow, left knee, and right knee are labeled in the 3D model;

[0076] N frames of images are obtained on average from the video data; where N is an integer greater than or equal to 2.

[0077] It should be further noted that the image at the reference time is the first frame, and the image at the last time is the last frame.

[0078] Each frame of the image is numbered according to the time sequence, and the number is represented by n; where n takes the value 1, 2...N;

[0079] The position coordinates of each frame of the image are obtained according to the spatial rectangular coordinate system; wherein, the position coordinates include the head coordinates, left elbow coordinates, right elbow coordinates, left knee coordinates, and right knee coordinates;

[0080] The head coordinates are marked as A. n (x n ,y n ,z n )

[0081] The coordinate of the left elbow is marked as B. 左n (x 左n ,y 左n ,z 左n )

[0082] The coordinate of the right elbow is marked as B. 右n (x 右n ,y 右n ,z 右n )

[0083] The coordinates of the left knee are marked as C. 左n (x 左n ,y 左n ,z 左n )

[0084] The coordinates of the right knee are marked as C. 右n (x 右n ,y 右n ,z 右n )

[0085] The position coordinates of each frame are combined into a coordinate group, resulting in a total of N coordinate groups.

[0086] Integrate the N sets of coordinates into the original data;

[0087] Obtain the violation detection model from the data processing module;

[0088] The raw data is input into the violation detection model to obtain the corresponding violation label;

[0089] The data processing module identifies the illegal labels;

[0090] When the violation label identifies an athlete as having committed a violation, the corresponding athlete's score will be cancelled.

[0091] In this embodiment, the violation label is either 0 or 1. When the violation label is 0, it indicates that the corresponding athlete has committed a violation, and when the violation label is 1, it indicates that the corresponding athlete has not committed a violation. In other preferred embodiments, the violation label can also be distinguished by other markers, such as the violation label being A or B. When the violation label is A, it indicates that the corresponding athlete has committed a violation, and when the violation label is B, it indicates that the corresponding athlete has not committed a violation.

[0092] In this embodiment, the violation detection model is built based on an artificial intelligence model, and the specific process includes:

[0093] Obtain standard training data from the data processing module;

[0094] The artificial intelligence model is trained using standard training data, and the trained artificial intelligence model is then marked as a violation detection model.

[0095] In this embodiment, the standard training data includes several sets of input data and corresponding violation labels, and the content attributes of the input data and the original data are consistent; it can be understood that both the input data and the original data include N selected coordinate groups, only the specific content of the coordinate groups is different.

[0096] In this embodiment, the artificial intelligence model includes models with strong nonlinear fitting capabilities, such as deep convolutional neural network models or RBF neural network models.

[0097] In this embodiment, the data acquisition module is communicatively and / or electrically connected to the data processing module.

[0098] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A computer vision-based running violation detection system, characterized in that, It includes a data acquisition module and a data processing module; the modules interact with each other using digital signals. The data acquisition module is used to acquire video data; wherein, the video data includes frontal video data, overhead video data, and side video data; And send the video data to the data processing module; The data processing module is used to receive the video data and obtain violation tags based on the video data and the violation detection model. The system also identifies the violation tags and disqualifies athletes from performing violations; the violation detection model is based on an artificial intelligence model. The data acquisition module acquires video data, specifically through the following process: The data acquisition module obtains the time of the gunshot and uses the time of the gunshot as the reference time. The data acquisition module acquires video data taken by the video acquisition device at the front position 0.1 seconds after the reference time, and marks it as the front video data; The data acquisition module acquires video data captured by the video acquisition device directly above at a reference time 0.1 seconds later, and marks it as the video data above. The data acquisition module acquires video data captured by the video acquisition device at the front and side positions 0.1 seconds after the reference time, and marks it as side video data; The data acquisition module sends the front video data, the top video data, and the side video data to the data processing module; The process of obtaining violation tags based on the video data and violation detection model includes: The data processing module receives the video data; Set the position of the reference object in the foreground video data as the origin, draw a straight line parallel to the horizontal plane through the origin, and this straight line is the x-axis. The straight line perpendicular to the x-axis through the origin is the z-axis, and establish the first rectangular coordinate system xoz. Set the position of the reference object in the video data above as the origin, establish a second rectangular coordinate system xoy, set the x-axis in the first rectangular coordinate system as the x-axis of the second rectangular coordinate system, and set the line passing through the origin and perpendicular to the x-axis as the y-axis. Set the position of the reference object in the side video data as the origin, establish a third rectangular coordinate system zoy, set the z-axis in the first rectangular coordinate system as the z-axis in the third rectangular coordinate system, and set the y-axis in the second rectangular coordinate system as the y-axis in the third rectangular coordinate system. Set the position of the reference object as the origin, and establish a spatial rectangular coordinate system based on the first rectangular coordinate system, the second rectangular coordinate system, and the third rectangular coordinate system; A three-dimensional model of the athlete is established based on the video data and the spatial rectangular coordinate system; The athlete's head, left elbow, right elbow, left knee, and right knee are labeled in the 3D model; N frames are obtained from the video data on average; where N is an integer greater than or equal to 2; where the image at the reference time is the first frame and the image at the last time is the last frame. Each frame of the image is numbered according to the time sequence, and the number is represented by n; where n takes the value 1, 2...N; The position coordinates of each frame of the image are obtained according to the spatial rectangular coordinate system; wherein, the position coordinates include the head coordinates, left elbow coordinates, right elbow coordinates, left knee coordinates, and right knee coordinates; The head coordinates are marked as A. n (x) n ,y n ,z n ) The coordinate of the left elbow is marked as B. 左n (x) 左n ,y 左n ,z 左n ) The coordinate of the right elbow is marked as B. 右n (x) 右n ,y 右n ,z 右n ) The coordinates of the left knee are marked as C. 左n (x) 左n ,y 左n ,z 左n ) The coordinates of the right knee are marked as C. 右n (x) 右n ,y 右n ,z 右n ) The position coordinates of each frame are combined into a coordinate group, resulting in a total of N coordinate groups. Integrate the N sets of coordinates into the original data; Obtain the violation detection model from the data processing module; The raw data is input into the violation detection model to obtain the corresponding violation label; The data processing module identifies the illegal labels; When the violation label is identified as indicating that an athlete has committed a violation, the corresponding athlete's score will be cancelled. The violation detection model is based on an artificial intelligence model, and the specific process includes: Obtain standard training data from the data processing module; The artificial intelligence model is trained using standard training data, and the trained artificial intelligence model is then labeled as a violation detection model; the artificial intelligence model includes a deep convolutional neural network model or an RBF neural network model.

2. The running violation detection system based on computer vision according to claim 1, characterized in that, The data acquisition module includes three video acquisition devices; The video capture devices are fixedly installed directly in front of, above, and to the side of the athlete's starting position.

3. The running violation detection system based on computer vision according to claim 1, characterized in that, The violation label can be either 0 or 1. When the violation label is 0, it means that the corresponding athlete has committed a violation. When the violation label is 1, it means that the corresponding athlete has not committed a violation.

4. The running violation detection system based on computer vision according to claim 1, characterized in that, The data acquisition module communicates with and / or is electrically connected to the data processing module.

Images