Intelligent shuttlecock serving judgment system based on computer vision and machine learning

Abstract

The invention relates to the technical field of shuttlecock serving judgment, and particularly discloses an intelligent shuttlecock serving judgment system based on computer vision and machine learning. An image processing module; the primary static judgment module is used for establishing a rule constraint model, inputting the judgment frame into the rule constraint model and judging whether the service has static violation or not; the secondary dynamic judgment module is used for judging whether dynamic violation exists in the serving process or not by continuously analyzing the motion trail under the condition that static violation does not exist; a result output module; the secondary dynamic judgment and optimization module is used for judging weak parts of the players and optimizing calculation of secondary dynamic violation evaluation values in the follow-up serving judgment process of the players; by analyzing the historical serving records of the players, the weak parts of the players are marked, and the weights are optimized in the subsequent serving judgment process, so that personalized serving judgment is realized, and the accuracy and fairness of serving judgment are further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of serving referees, and in particular to an intelligent badminton serving referee system based on computer vision and machine learning. Background Art

[0002] In the badminton competition system, the fairness and accuracy of serve penalties have always been the key link to ensure fair competition in the game. At present, the work of badminton serve referees mainly relies on manual decisions, but manual referees have exposed significant limitations when faced with complex and changeable serve scenarios.

[0003] In terms of the judgment method, traditional manual referees are more based on visual capture at the moment of serving, and only judge whether the serve is legal based on static features such as the player's posture and racket position at a specific moment. For example, when observing a player's serve, they only pay attention to whether the feet are in the serving area at the moment of serving, whether the angle between the racket face and the ground at the moment of hitting the ball complies with the rules, etc. However, this method is limited by the referee's vision range, reaction speed and subjective judgment differences. In some high-speed serving scenarios, the player's extremely subtle movement deviations, such as slight changes in the height of the racket head at the moment of serving, may be ignored due to the referee's untimely visual capture or poor angle, thus leading to misjudgment.

[0004] At the same time, existing refereeing methods lack dynamic analysis of the overall motion trajectory of the serving process and are unable to conduct a consistency review of the entire process from the serving preparation stage to the end of the serving. For example, players may deliberately cause abnormal interruptions or mutations in the motion trajectory during the serving process to confuse the referee, but traditional manual referees find it difficult to conduct in-depth analysis and accurate identification of the overall smoothness and consistency of the serving action.

[0005] With the rapid development of badminton, the pace of the game is accelerating, and the serving techniques are becoming increasingly diverse and complex, which puts higher demands on the accuracy and efficiency of the serving referees. Simply relying on traditional manual refereeing methods can no longer meet the strict standards of fair and just judgment in modern badminton events. Summary of the Invention

[0006] The object of the present invention is to provide an intelligent badminton serve referee system based on computer vision and machine learning to solve at least one of the above-mentioned problems in the prior art.

[0007] The present invention provides an intelligent badminton serve referee system based on computer vision and machine learning, which is characterized by comprising: Image acquisition module: acquire badminton serve images; Image processing module: Constructs the motion trajectory of the player, racket, and badminton; at the same time, extracts the judgment frame in the badminton serve image; A static judgment module: establishes a rule constraint model, inputs the judgment frame into the rule constraint model, and determines whether there is a static violation in the serve; Secondary dynamic judgment module: In the absence of static violations, the module extracts the complete motion trajectory of the player, racket, and shuttlecock during the serve process. By analyzing the continuity of the motion trajectory, it outputs a secondary dynamic violation assessment value to determine whether there is a dynamic violation during the serve process. Result output module: sends the determination results of the primary static determination module and the secondary dynamic determination module to the manual determination terminal, and generates a corresponding determination signal based on the results of the manual determination; Secondary dynamic judgment optimization module: Analyzes the stability of the player's parts in historical serve records, determines the player's weak points, and optimizes the calculation of secondary dynamic violation assessment values ​​in the subsequent player's serve refereeing process.

[0008] Beneficial effects of the present invention: 1. The present invention combines static and dynamic judgment to conduct a comprehensive and multi-level review of serves, greatly improving the accuracy of refereeing and providing a solid guarantee for the fairness of the game; 2. The present invention marks the weak points of players by analyzing their historical serving records, and optimizes the weights in the subsequent serving refereeing process, thereby realizing personalized serving judgment and further improving the accuracy and fairness of serving referees. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 This is a flow chart of an intelligent badminton serve referee system based on computer vision and machine learning provided in Example 1 of the present invention; Figure 2 1 is a schematic diagram of the structure of an intelligent badminton serve referee system based on computer vision and machine learning provided in Example 1 of the present invention; Figure 3 This is a structural diagram of an intelligent badminton serve referee device based on computer vision and machine learning provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0011] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0012] Example 1 Figure 1 The flowchart of the intelligent badminton serve referee system based on computer vision and machine learning provided in the first embodiment of the present invention is applicable to the case of an intelligent badminton serve referee based on computer vision and machine learning. The intelligent badminton serve referee system based on computer vision and machine learning can be executed by an intelligent badminton serve referee system based on computer vision and machine learning. The intelligent badminton serve referee system based on computer vision and machine learning can be implemented by software and / or hardware. The intelligent badminton serve referee system based on computer vision and machine learning can be configured in an intelligent badminton serve referee device based on computer vision and machine learning. Optionally, the intelligent badminton serve referee device based on computer vision and machine learning can be an electronic device, such as a laptop, desktop computer, or smart tablet, which is not limited by the embodiment of the present invention.

[0013] The intelligent badminton serve referee system based on computer vision and machine learning provided by the embodiment of the present invention specifically includes: Image acquisition module: using image acquisition equipment to obtain badminton serve images; Among them, image acquisition equipment includes but is not limited to: professional-grade high-resolution cameras; Image processing module: This module uses optical flow to process badminton serve image sequences frame by frame, calculating the motion vectors of pixels between adjacent frames to construct the motion trajectories of the player, racket, and shuttlecock. Furthermore, it incorporates a feature point tracking algorithm to continuously track key feature points on moving objects, accurately capturing changes in their motion state. Image frames with significant feature point changes are marked as decision frames. Among them, feature point tracking algorithms include but are not limited to: SIFT (Scale Invariant Feature Transform) or ORB (Accelerated Robust Features); For example, during a serve, the motion trajectory and speed changes of the racket head feature points are monitored in real time. When the racket is detected to have a speed exceeding a set threshold (e.g., 100 pixels per second) within a very short period of time (e.g., 0.05 seconds) and a significant change in motion direction occurs, the corresponding image frame will be preliminarily marked as a judgment frame. A static judgment module: Based on the badminton serve rules, a comprehensive and detailed rule constraint model is established. The judgment frame is input into the rule constraint model to determine whether the serve is illegal. Based on the judgment result, a static violation prompt signal and a static normal signal are generated; Exemplarily, the judgment frame is analyzed through a rule-constrained model. For the player's body posture, the rule-constrained model determines whether the feet remain within the serving area and do not leave the ground, and whether there are any illegal movements such as excessive twisting of the body by analyzing the position and angle of each joint of the player's body (such as shoulders, elbows, wrists, hips, knees, etc.); for the racket status, image recognition technology is used to accurately detect whether the position of the racket head is lower than the wrist at the moment of serving, and whether the angle between the racket face and the ground is within a specified range (such as an angle between 15° and 45° with the ground); if any of the above illegal movements exists, the serve is determined to be a suspected illegal serve, and a static illegal prompt signal is generated; if none of the above illegal movements exists, the serve is determined to be a normal serve, and a static normal signal is generated; Secondary dynamic judgment module: Based on the primary static normal signal, it extracts the complete motion trajectory of the player, racket, and badminton during the serve. By dynamically analyzing the continuity of the motion trajectory, it determines whether there are abnormal interruptions or mutations in the serve process. Based on the judgment results, it generates a secondary dynamic violation warning signal and a secondary dynamic normal signal. In some embodiments, based on a static normal signal, the complete motion trajectory of each part of the player's body (such as the arm, wrist), the racket, and the badminton from the serve preparation stage to the serve completion stage is extracted, and the smoothness index of each motion trajectory is calculated, and the smoothness value of each part of the player is output respectively. 、 、……、 , racket smoothness value PP, badminton smoothness value PY, where m is the total number of body parts of the player; It should be explained that smoothness is used to measure the stability and consistency of the motion trajectory. In a normal serve, the trajectory of the racket and other objects should be relatively smooth, that is, have a high smoothness value. When the smoothness is too low, it indicates that the motion trajectory is unstable, has sudden changes, or has abnormal interruptions. Such a serve is suspicious and may involve violations or other abnormal situations. If the smoothness is high, it usually means that the motion trajectory is relatively smooth and normal, and it is generally not marked as a suspicious serve. The process of obtaining the racket smoothness value PP is as follows: Based on a static normal signal, the complete racket motion trajectory from the serve preparation stage to the serve end stage is extracted, and the racket motion trajectory is represented by a series of coordinate point sequences to obtain the racket motion trajectory data as follows: , q is the total number of trajectory points in the racket trajectory; Normalize the racket trajectory data and map all coordinate values ​​to the [0,1] interval to eliminate the differences in the motion range of different moving objects; Exemplarily, the normalization method is minimum-maximum normalization, and its formula is: , where x is the original coordinate value, and are the minimum and maximum values ​​in the racket motion trajectory data, is the normalized racket motion trajectory data; Then, an autoencoder is preset. The autoencoder is an unsupervised learning model. The autoencoder includes an encoder and a decoder. The encoder compresses the input motion trajectory data into a low-dimensional potential representation, and the decoder reconstructs the potential representation into the original data. The normalized motion trajectory data is divided into a training set and a validation set. The autoencoder is trained using the training set. During the training process, the mean squared error (MSE) is used as the loss function. The model parameters are updated through stochastic gradient descent (SGD) or its variants (such as the Adam optimizer) to minimize the reconstruction error. After training, the model is evaluated using the validation set, and the model hyperparameters, such as the number of hidden layer units and the learning rate, are adjusted to improve the model performance. Apply the encoder part of the trained autoencoder to the racket motion trajectory data to obtain the racket's latent feature vector; At the same time, based on the judgment frame corresponding to a static normal signal, square grids of different sizes are defined to cover the plane where the racket's motion trajectory is located, and the number of grids containing motion trajectory points under each grid size is counted. , where e is the side length of the grid; For different values ​​of e, calculate and , and then fitted by linear regression and The relationship between the two, the slope is marked as the fractal dimension D of the racket trajectory; Then normalize the obtained fractal dimension so that it is in the range of [0,1]. Among them, the normalization processing method can be minimum-maximum normalization, and its formula is: ,in, and are the minimum and maximum values ​​of the fractal dimension in the trajectory of the racket respectively; The latent feature vector extracted by the autoencoder and the normalized fractal dimension are fused to output the racket smoothness value PP; For example, the latent feature vector extracted by the autoencoder and the normalized fractal dimension can be fused using a linear combination method, that is, ,in, is the latent feature vector extracted by the autoencoder, is the proportional coefficient, and its optimal value is determined by cross-validation and other methods; It should be explained that the smoothness value of the player part 、 、……、 The process of obtaining the badminton smoothness value PY is the same as that of obtaining the racket smoothness value PP, and will not be repeated here; By formula: , calculate the secondary dynamic violation evaluation value EG, where, 、 、 、 are all weight coefficients, and are all greater than 0, and , , is the player part smoothness value of the j-th part of the player, and the value of j is 1, 2, ..., m; Preset a secondary dynamic violation assessment threshold, and compare and analyze the secondary dynamic violation assessment value with the secondary dynamic violation assessment threshold; If the secondary dynamic violation evaluation value is less than or equal to the secondary dynamic violation evaluation threshold, the serve is determined to be an illegal serve, and a secondary dynamic violation prompt signal is generated; If the secondary dynamic violation evaluation value is greater than the secondary dynamic violation evaluation threshold, the serve is determined to be an illegal serve and a secondary dynamic normal signal is generated; Result output module: Based on the primary static violation prompt signal and the secondary dynamic violation prompt signal, the suspected violation data is sent to the manual judgment terminal, and based on the results of the manual judgment, a primary static violation signal, a primary static violation misjudgment signal, a secondary dynamic violation signal, and a secondary dynamic violation misjudgment signal are generated; Among them, suspected violation data includes but is not limited to: the judgment frame corresponding to the generation of a static violation prompt signal, the complete movement trajectory of the player's body parts (such as arms, wrists), rackets and badminton from the serve preparation stage to the serve end stage corresponding to the generation of a secondary dynamic violation prompt signal, the smoothness value of the player's parts 、 、……、 , racket smoothness value PP, badminton smoothness value PY; The technical solution of the embodiment of the present invention is mainly as follows: the embodiment of the present invention captures the serving image through a high-definition camera, uses the optical flow method to track the motion trajectory of the player, racket and badminton, combines feature point tracking algorithms such as SIFT or ORB to mark key frames, and makes a static judgment based on badminton rules to analyze the player's posture and racket status in the judgment frame to preliminarily judge the compliance of the serve. Subsequently, a secondary dynamic judgment extracts the complete motion trajectory, calculates the smoothness index, uses the autoencoder to extract the feature vector and calculate the fractal dimension, comprehensively evaluates the smoothness and continuity of the serving action, and further determines the possibility of violation. The suspected violation data is pushed to the manual judgment terminal, combined with manual review, and finally generates an accurate judgment signal. The present invention reviews the serve in an all-round and multi-level manner by combining static and dynamic judgment, which greatly improves the accuracy of the referee and provides a solid guarantee for the fairness of the game.

[0014] Example 2 Based on Example 1, please refer to Figure 1 、 Figure 3 As shown, the intelligent badminton serve referee system based on computer vision and machine learning according to the embodiment of the present invention further includes: Secondary dynamic judgment optimization module: Analyzes the stability of players' parts in historical serve records, identifies players' weak points, and optimizes the calculation of secondary dynamic violation assessment values ​​in subsequent players' serve refereeing processes; Based on any player, calculate the position stability value of all player position smoothness values ​​at different serve times in the player's historical serve records. The specific process is: Extracting the player's part smoothness values ​​at different serve times in the player's historical serve records, presetting the corresponding player part smoothness threshold, and comparing and analyzing the player part smoothness values ​​with the player part smoothness threshold; If the player part smoothness value is less than or equal to the player part smoothness threshold, it means that the player part has an abnormality in the number of serves corresponding to the player part smoothness value, and a serve part abnormality signal is generated; If the player part smoothness value is greater than the player part smoothness threshold, it means that there is no abnormality in the part of the player in the number of serves corresponding to the player part smoothness value, and a normal serve part signal is generated; Extract the number of abnormal signals generated from all serving positions, calculate the ratio with the total number of serves by the player in the historical serving records, and output the player's standard abnormality ratio; At the same time, based on the smoothness values ​​of the player's parts at different serves in the player's historical serve records, the first quartile, third quartile, maximum value and minimum value are extracted, and the difference between the third quartile and the first quartile is calculated to output the player's own stable value. The difference between the maximum value and the minimum value is calculated to obtain the player's own floating value; Calculate the ratio of the player's own stable value to the player's own floating value, and output the player's own abnormality ratio; Then, the player's standard abnormality ratio and the player's own abnormality ratio are weighted and summed to output the position stability value; Preset the part stability threshold, and compare and analyze the part stability value with the part stability threshold; If the part stability value is less than or equal to the part stability threshold, it means that the motion trajectory of the body part corresponding to the part stability value is relatively stable when the player serves, and the body part is marked as a stable part; If the part stability value is greater than the part stability threshold, it means that the movement trajectory of the body part corresponding to the part stability value of the player is relatively unstable when serving, and the body part is marked as a weak part; It should be explained that weak areas refer to areas where players have difficulty maintaining stable and coherent movements when serving, and are therefore prone to small movements. Therefore, weak areas will be given special attention in future serve decisions to reduce misjudgments caused by small movements. Based on the players' weak points, the calculation of the secondary dynamic violation assessment value during the subsequent player's serve refereeing process is optimized. The specific process is as follows: When calculating the secondary dynamic violation assessment value, the weight corresponding to the weak part is multiplied by the optimization coefficient to obtain the new weight of the weak part; The optimization coefficient is calculated as follows: the ratio of the stability value of the weak part to the stability threshold is calculated to obtain the optimization coefficient; The technical solution of the embodiment of the present invention is mainly as follows: the embodiment of the present invention adds a secondary dynamic judgment optimization module, which analyzes the stability of the player's parts in the historical serving records, determines the player's weak parts, and optimizes the calculation of the secondary dynamic violation assessment value in the subsequent serving referee process. Specifically, the module extracts the player's part smoothness value in the historical serving records, and calculates the part stability value and the player's own abnormality ratio, obtains the part stability value through weighted summation, compares the part stability value with the preset part stability threshold, marks the stable parts and weak parts, and in future serving judgments, pays special attention to the weak parts and adjusts their corresponding weights to reduce misjudgments caused by small movements.

[0015] Example 3 Reference Figure 3 An embodiment of the present invention further provides a computer device 3, comprising: a memory 302, a processor 301, and a computer program 303 stored in the memory 302. When the computer program 303 is executed on the processor 301, an intelligent badminton serve referee system based on computer vision and machine learning as described in any one of the above methods is implemented.

[0016] The computer device 3 may be a desktop computer, a notebook computer, a PDA, a cloud server or other computing devices. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that Figure 3 This is merely an example of the computer device 3 and does not constitute a limitation on the computer device 3 . The computer device 3 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 3 may also include input and output devices, network access devices, etc.

[0017] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0018] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard drive or memory of the computer device 3. In other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the computer device 3. Furthermore, the memory 302 may include both an internal storage unit of the computer device 3 and an external storage device. The memory 302 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 302 may also be used to temporarily store data that has been output or is about to be output.

[0019] Example 4 An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computer vision and machine learning-based intelligent badminton serve referee system as described in any one of the above methods is implemented.

[0020] In this embodiment, if the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a camera / terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, based on legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0021] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0022] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0023] In the embodiments disclosed in the present application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely schematic. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0024] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0025] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0026] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An intelligent badminton serve referee system based on computer vision and machine learning, characterized by: include: Image acquisition module: acquire badminton serve images; Image processing module: Constructs the motion trajectory of the player, racket, and badminton; at the same time, extracts the judgment frame in the badminton serve image; A static judgment module: establishes a rule constraint model, inputs the judgment frame into the rule constraint model, and determines whether there is a static violation in the serve; Secondary dynamic judgment module: In the absence of static violations, the module extracts the complete motion trajectory of the player, racket, and shuttlecock during the serve process. By analyzing the continuity of the motion trajectory, it outputs a secondary dynamic violation assessment value to determine whether there is a dynamic violation during the serve process. Result output module: sends the determination results of the primary static determination module and the secondary dynamic determination module to the manual determination terminal, and generates a corresponding determination signal based on the results of the manual determination; Secondary dynamic judgment optimization module: Analyzes the stability of the player's parts in historical serve records, determines the player's weak points, and optimizes the calculation of secondary dynamic violation assessment values ​​in the subsequent player's serve refereeing process.

2. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 1, characterized in that: The process of inputting the judgment frame into the rule constraint model to determine whether there is a static violation in the serve is as follows: The rule constraint model determines whether there is an illegal action by analyzing the position and angle of each joint of the player's body and the position of the racket head at the moment of serving. If there is an illegal action, the serve is judged to be an illegal serve and a violation judgment signal is generated. If there is no illegal action, the serve is judged to be a normal serve and a static normal signal is generated.

3. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 2, characterized in that: By analyzing the continuity of the motion trajectory and outputting the secondary dynamic violation evaluation value, the process of determining whether there is a dynamic violation in the serving process is as follows: Based on a static normal signal, the complete motion trajectory of each part of the player's body, the racket, and the badminton from the serve preparation stage to the serve completion stage is extracted, and the smoothness index of each motion trajectory is calculated. The smoothness index of the player's body part, racket smoothness value, and badminton smoothness value are output respectively. After performing a weighted summation, the secondary dynamic violation assessment value is output; If the secondary dynamic violation assessment value is less than or equal to the secondary dynamic violation assessment threshold, a secondary dynamic violation prompt signal is generated; If the secondary dynamic violation assessment value is greater than the secondary dynamic violation assessment threshold, a secondary dynamic normal signal is generated.

4. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 3, characterized in that: The process of obtaining the racket smoothness value is: Based on a static normal signal, the complete racket motion trajectory data from the serve preparation stage to the serve completion stage is extracted; Normalize the racket motion trajectory data and output the normalized racket motion trajectory data ; The racket motion trajectory data after normalization , respectively obtain the latent feature vector and the normalized fractal dimension, fuse them, and output the racket smoothness value.

5. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 4, characterized in that: The process of obtaining the potential feature vector is: Preset autoencoder, which includes encoder and decoder; The normalized motion trajectory data is divided into a training set and a validation set. The training set is used to train the autoencoder. After the training is completed, the validation set is used to evaluate the model. The encoder part of the trained autoencoder is applied to the racket motion trajectory data to obtain the racket's latent feature vector.

6. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 4, characterized in that: The process of obtaining the normalized fractal dimension is: Based on the judgment frame corresponding to a static normal signal, define square grids of different sizes to cover the plane where the racket's motion trajectory is located, and count the number of grids containing motion trajectory points under each grid size , where e is the side length of the grid; For different values ​​of e, calculate and , and then fitted by linear regression and The relationship between the two, the slope is marked as the fractal dimension D of the racket trajectory; Then the obtained fractal dimension is normalized to obtain the normalized fractal dimension .

7. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 1, characterized in that: The process of sending the determination results of the primary static determination module and the secondary dynamic determination module to the manual determination terminal and generating the corresponding determination signal based on the manual determination results is as follows: Based on the primary static violation prompt signal and the secondary dynamic violation prompt signal, the suspected violation data is sent to the manual judgment terminal, and based on the result of the manual judgment, a primary static violation signal, a primary static violation misjudgment signal, a secondary dynamic violation signal, and a secondary dynamic violation misjudgment signal are generated; Among them, suspected violation data includes but is not limited to: the judgment frame corresponding to the generation of a static violation prompt signal, the complete movement trajectory of the player's body parts, racket and badminton from the serve preparation stage to the serve end stage corresponding to the generation of a secondary dynamic violation prompt signal, the smoothness value of the player's parts, the smoothness value of the racket, and the smoothness value of the badminton.

8. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 1, characterized in that: The process of obtaining weak points is as follows: Calculate the position stability value of all players' position smoothness values ​​at different serve times in the players' historical serve records; If the part stability value is greater than the part stability threshold, the body part is marked as a weak part.

9. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 8, characterized in that: The process of obtaining the part stability value is as follows: Extracting the player's part smoothness values ​​at different serve times in the player's historical serve records, presetting the corresponding player part smoothness threshold, and comparing and analyzing the player part smoothness values ​​with the player part smoothness threshold; If the player part smoothness value is less than or equal to the player part smoothness threshold, a serving part abnormality signal is generated; Extract the number of abnormal signals generated from all serving positions, calculate the ratio with the total number of serves by the player in the historical serving records, and output the player's standard abnormality ratio; At the same time, based on the smoothness values ​​of the player's parts at different serves in the player's historical serve records, the first quartile, third quartile, maximum value and minimum value are extracted, and the difference between the third quartile and the first quartile is calculated to output the player's own stable value. The difference between the maximum value and the minimum value is calculated to obtain the player's own floating value; Calculate the ratio of the player's own stable value to the player's own floating value, and output the player's own abnormality ratio; Then, the player's standard abnormality ratio and the player's own abnormality ratio are weighted and summed to output the position stability value.

10. The intelligent badminton serve referee system based on computer vision and machine learning according to claim 1, characterized in that: The process of optimizing the calculation of the secondary dynamic violation evaluation value in the subsequent player's serve refereeing process is as follows: When calculating the secondary dynamic violation assessment value, the weight corresponding to the weak part is multiplied by the optimization coefficient to obtain the new weight of the weak part; The calculation process of the optimization coefficient is as follows: the ratio of the part stability value corresponding to the weak part to the part stability threshold is calculated to obtain the optimization coefficient.

Images