Snooker artificial intelligence scoring system based on machine vision identification

Abstract

The invention relates to the technical field of billiards, in particular to a snooker artificial intelligence scoring system based on machine vision recognition, which comprises a video recording module, a score processing module, a posture acquisition module, a posture scoring module and a display module. The invention aims to provide the snooker artificial intelligence scoring system based on machine visual identification, and through deep learning and multi-sensor data fusion technology and in combination with deep fusion of batting postures of players, timeliness and objectivity in a snooker match are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of billiards, and particularly to a snooker artificial intelligence scoring system based on machine vision recognition. Background Art

[0002] The existing referee system for snooker competitions mainly relies on manual scoring and visual judgment, which has problems such as subjective judgment errors and reaction delays, and it is difficult to maintain stable and accurate sphere positioning and motion tracking in high-speed movement and complex lighting environments; in addition, traditional technologies are difficult to achieve comprehensive and real-time analysis of players' hitting actions, which limits the fairness of the competition and the objectivity of referee decisions. Summary of the Invention

[0003] In order to overcome the disadvantages and deficiencies existing in the prior art, the purpose of the present invention is to provide a snooker artificial intelligence scoring system based on machine vision recognition, which effectively improves the timeliness and objectivity in snooker competitions through deep learning and multi-sensor data fusion technology, combined with the deep fusion of players' hitting postures.

[0004] The present invention is realized through the following technical solutions: In the first aspect, the present invention discloses a snooker artificial intelligence scoring system based on machine vision recognition, including: A video recording module, configured to acquire video information of the spheres on the billiard table and send the video information to the processing module; A score processing module, configured to identify the position, size, and category of the spheres in each frame of the video information, track the movement trajectory of each sphere, and judge the legality of the hit and calculate the score for the hit according to the snooker rules; A posture acquisition module, configured to acquire the human body posture image of the player during the hit and output the key part feature point data of the player's body; A posture scoring module, configured to receive the feature point data, calculate the index of the feature points according to the standard feature points combined with a mathematical model, quantitatively evaluate the standardization of the hitting action, and output the posture score of the player's hit; And a display module, configured to display the hit score and the posture score of the current player.

[0005] In the second aspect, the present invention discloses a snooker artificial intelligence scoring method based on machine vision recognition, and the method includes the following steps: S100. Use a neural convolutional network to perform target detection on the spheres, combine a multi-object tracking algorithm to achieve real-time positioning and tracking of the spheres on the table, and record the trajectories, positions, and movement speeds of each sphere; S200. Obtain the human body posture image and the cue image of the player when hitting the ball, and output the posture score of the player's shot according to the key part feature point data of the player's body and the feature point data of the cue; S300. Use the real-time detected sphere trajectory, position and movement speed, and combine the snooker rules and physical models to calculate the collision time between the balls and judge the legality of the shot; S400. Dynamically update the scores of both sides according to the scoring rules of each ball and the shooting order, and perform special deduction processing when a foul occurs; S500. Based on the output of the foregoing steps, use the fusion decision algorithm to make the final referee decision.

[0006] Combined with the second aspect, further, in step S100, the target detection of the sphere includes the following steps: Define a detection network function: Wherein, is the set of detection results, is the input image, is the set of model parameters, is the th detection result, including the center position , width , height , confidence and the category of the ball , is the number of candidate detections; Establish a state vector for each detected sphere: Wherein, is the time The position coordinates of the sphere in the image or the actual coordinate system, is the time The sphere is at and The speed in the direction; Establish the prediction and update steps of the sphere. The prediction step includes: State prediction: Covariance prediction: Wherein, is the predicted state estimate, is the state transition matrix, is the posterior state estimate, is the control matrix, is the process noise covariance matrix, is the state covariance predicted to the current moment through state transition, is the uncertainty of the state estimate after measurement update at the previous moment, is the state transition matrix is the transpose matrix of; The update steps include: Kalman gain calculation: State update: Covariance update: where, is the observation matrix, is the measurement noise covariance matrix, is the Kalman gain, is at time the predicted observation value, is the identity matrix.

[0007] Combined with the second aspect, further, in step S200, the calculation steps of the posture score of the player's batting include: Extract the human body set through the deep neural network: where, is the coordinate of the th key point in the image, is the confidence of the th key point, is the number of detected key points; Combined with multi-view images, use triangulation to recover the 3D coordinates. Let the corresponding 2D points obtained from two cameras be and , then solve the three-dimensional position and through the projection matrices satisfies: Calculate the bending degree of the joint: where, are the 3D coordinates of the corresponding joint points respectively; Model the bending degree of the joint through the time series model, and construct the LSTM update formula to output the posture score of the player's batting.

[0008] In combination with the second aspect, further, in step S300, the steps for judging the legality of a shot include: Set the motion state of the white ball as , and the state of the target candidate ball as , and predict the collision time of the spheres: Among them, is the initial displacement vector, is the relative velocity, and are the radii of the white ball and the target ball respectively; Define the judgment function for a legal shot: Among them, is the set of all candidate balls in this round, is the time when the white ball and the th target ball predict a collision when moving in a straight line according to the current motion state; is the collision time corresponding to the ball that the white ball predicts to collide with first among all candidate target balls; is the category of the first colliding ball, is the category of the target ball specified in the current round.

[0009] In combination with the second aspect, further, in step S400, the steps for updating the scores of both sides include: Define the score status update function: Among them, is the score at time , is the number of legally pocketed balls during this shot, is the indicator function, is the ball corresponding score; Regard the score status update function as a turntable transfer equation: Among them, is the score of the corresponding ball, is the set of rule parameters, is a piecewise function based on rules or a decision tree model integrating expert knowledge.

[0010] In combination with the second aspect, further, in step S500, the steps for the final referee's judgment include: Define the decision feature vector: Among them, is the confidence level of a legal shot, is the score change, is the position and velocity information of the ball, " " is the scoring of the batting posture; A multi-classifier is used for the final decision: Among them, is the high-order feature fusion function, is the network weight, is the probability that the referee's decision belongs to class ; The final decision is: Thirdly, the present invention also discloses an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor runs the computer program to implement the above-mentioned snooker artificial intelligence scoring method based on machine vision recognition.

[0011] Fourthly, the present invention also discloses a computer-readable storage medium, on which a computer program is stored. The program is executed by the processor to implement the above-mentioned snooker artificial intelligence scoring method based on machine vision recognition.

[0012] Advantages of the present invention: A snooker artificial intelligence scoring system, method, electronic device and storage medium of the present invention adopt deep learning and multi-sensor data fusion technology to achieve high-precision real-time positioning and tracking of the balls on the table, as well as 3D reconstruction and time-series analysis of the batting postures of the players, thereby significantly improving the accuracy and fairness of the game scoring and referee decisions; in addition, through physical models and high-order feature fusion algorithms, intelligent analysis is carried out on the collision time of the white ball and the target ball, the legality of the shot, and the actions of the players, reducing human misjudgments and improving the operation efficiency of the game and the objectivity of on-site decisions. Description of the Drawings

[0013] The present invention is further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to the following drawings without creative efforts.

[0014] Figure 1 is the step flowchart of the snooker artificial intelligence scoring method provided by the embodiment of the present invention. Detailed Embodiments

[0015] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the accompanying drawings. Many specific details are set forth in the following description to facilitate a thorough understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0016] Existing snooker game referee systems mainly rely on manual scoring and visual judgment, suffering from problems such as subjective judgment errors and reaction delays, and it is difficult to maintain stable and accurate ball positioning and motion tracking in high-speed movement and complex lighting environments. In addition, traditional technologies are difficult to achieve comprehensive and real-time analysis of players' hitting actions, restricting the fairness of the game and the objectivity of referee decisions.

[0017] To solve the above problems, this embodiment discloses a snooker artificial intelligence scoring system based on machine vision recognition, which includes: A video recording module for acquiring video information of the balls on the billiard table and sending the video information to the processing module; A score processing module for identifying the position, size, and category of the balls in each frame of the video information, tracking the motion trajectories of each ball, and judging the legality of the hit and calculating the score for the hit according to snooker rules; A posture acquisition module for acquiring the human body posture image of the player during the hit and outputting the key part feature point data of the player's body; A posture scoring module for receiving the feature point data, calculating the metrics of the feature points according to the standard feature points combined with a mathematical model, quantitatively evaluating the standardization of the hitting action and outputting the posture score of the player's hit; And a display module for displaying the hit score and the posture score of the current player.

[0018] In this embodiment, the video recording module and the posture acquisition module are preferably high-definition cameras or high-definition video cameras, the score processing module and the posture scoring module are preferably computers equipped with computing servers, and the display module is preferably a screen.

[0019] Please refer to Figure 1 , which shows a snooker artificial intelligence scoring method provided by some embodiments of the present application. This method can be executed on the system as described above, and this method includes the following steps: S100. Use a neural convolutional network to perform object detection on the balls, and combine a multi-object tracking algorithm to achieve real-time positioning and tracking of the balls on the table, and record the trajectories, positions, and movement speeds of each ball; S200. Obtain the human body posture image and the cue image of the player when hitting the ball, and output the posture score of the player's hitting according to the key part feature point data of the player's body and the feature point data of the cue; S300. Use the real-time detected sphere trajectory, position and movement speed, and combine the snooker rules and physical models to calculate the collision time between the balls and judge the legality of the hit; S400. Dynamically update the scores of both sides according to the scoring rules of each ball and the hitting order, and perform special deduction processing when a foul occurs; S500. Integrate the outputs of the foregoing steps and use the fusion decision algorithm to make the final referee decision.

[0020] In step S100, the steps for target detection of the sphere include the following: Define a detection network function: Among them, is the set of detection results, is the input image. Here, can be a frame of video or a static picture, containing all the visual information in the scene; is the set of model parameters, usually including the weights, biases and other parameters to be trained in the convolutional neural network (CNN). These parameters are learned through a large amount of training data, enabling the model to effectively extract features in the image and complete the target detection task; is the th detection result, containing the center position width height confidence and the category of the ball , is the number of candidate detections; Establish a state vector for each detected sphere: Among them, is the time the position coordinates of the sphere in the image or the actual coordinate system, is the time the sphere is at and the speed in the direction; Establish the prediction and update steps of the sphere. The prediction steps include: State prediction: Covariance prediction: Among them, is the predicted state estimate, is the state transition matrix. In a uniform motion model, it is usually taken as: Among them, represents the time interval between adjacent frames, which is used to transfer the state at the previous moment to the predicted state at the current moment. For example, the position of the sphere is obtained by adding the position at the previous moment to the product of the velocity and the time interval; is the posterior state estimate, is the control matrix. In most motion tracking problems, especially when purely observing the motion of the sphere, there is usually no explicit control input. In this case, can be set or this part can be directly ignored; is the process noise covariance matrix, is the state covariance predicted to the current moment through state transition, is the uncertainty of the state estimate updated through measurement at the previous moment, is the state transition matrix and is the transpose matrix of Using can ensure that the predicted covariance matrix Kalman gain calculation: State update: Covariance update: Among them, is the observation matrix. If the measurement directly gives the position of the sphere, then it is usually taken as: which is used to map the content in the state vector to the observation space. Since the measurement usually only contains position data, only extracts in and components; To measure the noise covariance matrix, which describes the uncertainty or noise existing in the measurement process (e.g., errors caused by lighting, blurring, or algorithm limitations during image detection), determines the respective weights of the detection result and the prediction result in the update step. The greater the noise, the lower the trust of the system in the measurement result, and it relies more on the prediction during update; is the Kalman gain, which determines the balance between the predicted state and the actual measurement data in the update step. It indicates how much weight should be given to the measurement data when the system corrects the prediction. A higher means relying more on the new measurement; a lower means having more trust in the prediction; is the predicted observation value at time , usually the two-dimensional position of the sphere; is the identity matrix, which plays a role in maintaining the original dimension and eliminating part of the predicted covariance in matrix operations.

[0021] The principle and effect of sphere object detection are explained below.

[0022] Sphere positioning (detection): Use a deep convolutional neural network (CNN) to process each frame of the image captured by the camera. The network can identify the positions and categories of each sphere on the tabletop (such as red balls, white balls, etc.). Specifically, the detection network outputs the bounding box (including the center coordinates, width, and height) and confidence score of each ball, and then removes duplicate or overlapping detection results through non-maximum suppression (NMS) to obtain the accurate positions of the spheres.

[0023] Sphere tracking: After detecting the spheres, a multi-object tracking algorithm (such as data association combining Kalman filtering and the Hungarian algorithm) is used to track the movement of each ball in consecutive frames. The Kalman filter predicts the state of the current frame based on the state (position and velocity) of the sphere in the previous frame and corrects it using the current detection result; while the Hungarian algorithm is used to establish the best matching relationship between multiple detection results and prediction results to ensure that each sphere can be accurately identified and tracked throughout the video sequence.

[0024] Furthermore, in step S200, the calculation steps of the player's batting posture score include: Extract the human body set through a deep neural network: Among them, is the coordinate of the th key point in the image, is for the The confidence of a key point is the number of detected key points; Combining multi-view images, use triangulation to recover 3D coordinates. Let the corresponding 2D points obtained from two cameras be and , then solve for the three-dimensional position and through the projection matrices satisfying: where, is the scale factor, representing the scaling factor during the projection process from 3D space to 2D image. Since the imaging process of the camera is usually perspective projection, the image coordinates of each 3D point will be affected by the relative position, angle, and distance between the camera and the object. The scale factor scales the coordinates in 3D space to the appropriate 2D image coordinate system; and are the two-dimensional coordinates of the th feature point in the image captured by the th camera. These are the pixel coordinates of the corresponding feature points in the 2D image captured by the camera. For example, if the camera captures the same object from two different angles, the two cameras will project different two-dimensional coordinates for the same object; is the projection matrix of the camera, which contains the internal parameters of the camera (such as focal length and principal point position) and external parameters (such as rotation matrix and displacement vector). Usually, these positions are represented in the camera coordinate system and then converted to the image coordinate system using the projection matrix.

[0025] The purpose of recovering 3D coordinates is to restore the 2D key points extracted from the image to the position coordinates in three-dimensional space. This process mainly depends on multi-view data and the internal and external parameters of the camera; Calculate the bending degree of the joint: where, are the 3D coordinates of the corresponding joint points respectively; Model the bending degree of the joint through a temporal model and construct an LSTM update formula to output the posture score of the player's batting.

[0026] Furthermore, in step S300, the judgment steps for the legality of batting include: Set the motion state of the white ball to , and the state of the target candidate ball to , predict the collision time of the spheres: Among them, is the initial displacement vector, is the relative velocity, and are the radii of the white ball and the target ball respectively; Define a judgment function for a legal shot: Among them, is the set of candidate balls in this round, is the time when the white ball and the th target ball predict a collision when moving in a straight line according to the current motion state; is the collision time corresponding to the ball that the white ball predicts to collide with first among all candidate target balls; is the category of the first colliding ball, is the category of the target ball specified in this round.

[0027] Furthermore, in step S400, the steps for updating the scores of both sides include: Define a score status update function: Among them, is the score at time , is the number of legally pocketed balls in this shot process, is an indicator function, is the ball corresponding score; Regard the score status update function as a turntable transfer equation: Among them, is the score of the corresponding ball, is the rule parameter set, is a piecewise function based on rules or a decision tree model integrating expert knowledge.

[0028] Furthermore, in step S500, the steps for the final referee's judgment include: Define a decision feature vector: Among them, is the confidence of a legal shot, is the score change, is the position and velocity information of the sphere, " " is the scoring of the batting posture; Adopt a multi-classifier for the final adjudication: Among them, is a high-order feature fusion function, is the network weight, is the probability that the referee's decision belongs to category ; The final decision is: In this embodiment, the calculation formula of the high-order feature fusion function is: Among them, and are the weights and biases of the neural network, responsible for mapping the input feature to a new feature space; is an activation function (such as ReLU or Sigmoid), used to introduce non-linearity; and are the weights and biases of the final output layer, which map the fused features to each possible referee decision category (for example: legal shot, foul, re-rack, etc.); The function is used to normalize the output value into a probability distribution, so that the output of each decision category is a probability value, representing the probability that the current action belongs to a certain category.

[0029] The output of

[0030] is a vector, and each element in it corresponds to the probability of a referee decision category.

[0031] Some embodiments of the present application also provide an electronic device, which includes: a processor, a memory, a bus, and a communication interface. The processor, the communication interface, and the memory are connected through the bus; a computer program that can run on the processor is stored in the memory, and when the processor runs the computer program, it executes the method provided in any of the foregoing embodiments of the present application.

[0032] Among them, the memory may include a high-speed random access memory (RAM: Random Access Memory), and may also include a non-volatile memory, such as at least one disk memory. Through at least one communication interface (which can be wired or wireless), a communication connection is realized between this system network element and at least one other network element, and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0033] The bus can be an ISA bus, a PCI bus, an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. Among them, the memory is used to store programs. After receiving an execution instruction, the processor executes the program. Any implementation manner of the disclosed Snooker artificial intelligence scoring method in the embodiments of the present application can be applied to or implemented by the processor.

[0034] The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit in the hardware of the processor or by instructions in the form of software. The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as being executed by the hardware decoding processor, or executed by a combination of the hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method.

[0035] The electronic device provided by the embodiments of the present application and the intelligent control method for heavy object handling provided by the embodiments of the present application are based on the same application concept and have the same beneficial effects as the method adopted, run, or implemented by it.

[0036] The embodiments of the present application also provide a computer-readable storage medium corresponding to the intelligent control method for heavy object handling provided by the foregoing embodiments. A computer program is stored thereon. The computer-readable storage medium is an optical disc, and a computer program (i.e., a program product) is stored thereon. When the computer program is run by the processor, it will execute the intelligent control method for heavy object handling provided by any of the foregoing embodiments.

[0037] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other optical and magnetic storage media, which will not be elaborated here one by one.

[0038] The computer-readable storage medium provided by the above embodiments of the present application and the snooker artificial intelligence scoring method provided by the embodiments of the present application are based on the same application concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0039] It should be noted that in the above text, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device including a series of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the presence of additional identical elements in the process, method, article or device including that element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in a reverse order according to the functions involved. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0040] From the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc) and includes several instructions for causing a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present application.

[0041] The embodiments of the present application have been described above in conjunction with the accompanying drawings, which are only specific implementation manners of the present application. However, the present application is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them belong to the protection scope of the present application.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting the protection scope of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A snooker artificial intelligence scoring system based on machine vision recognition, characterized in that: include: The video recording module is used to obtain the video information of the ball on the tabletop and send the video information to the processing module; The score processing module is used to identify the position, size and type of the ball in each frame of the video information, track the movement trajectory of each ball, and determine the legality of the shot according to the snooker rules and calculate the score of the shot; The posture acquisition module is used to acquire the human posture image of the player when hitting the ball, and output the key parts of the player's body feature point data; The posture scoring module is used to receive feature point data, calculate the index of the feature points based on the standard feature points combined with the mathematical model, quantitatively evaluate the standardization of the batting action and output the player's batting posture score; and a display module for displaying the current player's batting score and posture score; The scoring system is also used to perform the following steps: S100. Use a neural convolutional network to detect spheres, and combine it with a multi-eye tracking algorithm to achieve real-time positioning and tracking of spheres on the desktop, and record the trajectory, position, and movement speed of each sphere; S200. Obtaining the human body posture image and the club image of the player when hitting the ball, and outputting the player's hitting posture score according to the key feature point data of the player's body and the feature point data of the club; S300. Using the ball trajectory, position and movement speed detected in real time, combined with snooker rules and physical models, calculate the collision time between the balls to determine the legality of the shot; S400. Dynamically update the scores of both parties according to the scoring rules of each ball and the batting order, and make special point deductions when a foul occurs; S500. Combining the outputs of the previous steps, a fusion decision algorithm is used to make a final refereeing decision.

2. A snooker artificial intelligence scoring system based on machine vision recognition according to claim 1, characterized in that: In step S100, the sphere object detection includes the following steps: Define a detection network function: in, is the set of test results, is the input image, is the parameter set of the model, For the detection results, including the center position ,width ,high , Confidence And the type of ball , is the number of candidate tests; Create a state vector for each detected sphere: in, For time The position coordinates of the sphere in the image or real coordinate system, For time The sphere is and Direction speed; Establish the prediction and update steps of the sphere. The prediction step includes: Status prediction: Covariance prediction: in, To predict the state estimate, is the state transfer matrix, is the posterior state estimate, is the control matrix, is the process noise covariance matrix, is the state covariance predicted to the current moment through state transition, is the uncertainty of the state estimate after the measurement update at the last moment, is the state transfer matrix The transposed matrix of The update steps include: Kalman gain calculation: Status Update: Covariance update: in, is the observation matrix, is the measurement noise covariance matrix, is the Kalman gain, For in time The predicted observed value of is the identity matrix.

3. A snooker artificial intelligence scoring system based on machine vision recognition according to claim 1, characterized in that: In step S200, the steps of calculating the player's batting posture score include: Extract the human body collection through deep neural network: in, For the The coordinates of the key points in the image, For the first The confidence of the key points, is the number of key points detected; Combining multi-view images, the 3D coordinates are restored using triangulation. The corresponding 2D points obtained from the two cameras are assumed to be and , then through the projection matrix and Solving for 3D position satisfy: in, is the scale factor, and Respectively The feature points are The two-dimensional coordinates in the image taken by the space camera, is the projection matrix of the camera; Calculate the degree of flexion of a joint: in, are the 3D coordinates of the corresponding joint points respectively; The bending degree of the joints is modeled through a time series model, and an LSTM update formula is constructed to output the player's batting posture score.

4. A snooker artificial intelligence scoring system based on machine vision recognition according to claim 1, characterized in that: In step S300, the step of judging the legality of the shot includes: Set the motion state of the white ball to , the state of the target candidate ball is , predict the collision time of the spheres: in, is the initial displacement vector, is the relative speed, and The radius of the white ball and the target ball respectively; Define the judgment function for legal shots: in, For all candidate balls in this round, For the white ball and When the target balls move in a straight line according to the current motion state, predict the time when the collision will occur; Predict the collision time of the first ball that the white ball will collide with among all the candidate target balls; is the category of the first collision ball, The type of object ball specified for the current round.

5. A snooker artificial intelligence scoring system based on machine vision recognition according to claim 1, characterized in that: In step S400, the step of updating the scores of both parties includes: Define the scoring status update function: in, For time The necessity of time, is the number of balls legally pocketed during this shot, is the indicator function, For the ball The corresponding score; The scoring state update function can be viewed as a turntable transfer equation: in, is the score of the corresponding ball, is the rule parameter set, It is a rule-based piecewise function or a decision tree model that integrates expert knowledge.

6. A snooker artificial intelligence scoring system based on machine vision recognition according to claim 1, characterized in that: In step S500, the final referee determination step includes: Define the decision feature vector: in, is the confidence level of a legal shot, For the score change, is the position and velocity information of the sphere, ” Score your batting stance; Use a multi-classifier to make the final decision: in, is a high-order feature fusion function, is the network weight, For referee decisions belong to the category The probability of The final decision is:

7. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor runs the computer program to implement the steps executed in the snooker artificial intelligence scoring system based on machine vision recognition as described in any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by the processor to implement the steps performed in the snooker artificial intelligence scoring system based on machine vision recognition as described in any one of claims 1-6.

Images