System and method for simulating virtual players based on player weaknesses for targeted training

Abstract

A system and method for simulating a virtual player according to a player's weaknesses for targeted training, which generates weakness information of a target player's on-field behavior using a defensive behavior model generated after training an offensive behavior model and a defensive behavior model according to historical game videos, generates defensive behavior information according to the weakness information, generates a virtual opponent executing the defensive behavior information and provides interaction with the virtual opponent in a real-time scenario, adjusts player parameters of the target player when a training recommendation condition is satisfied according to the interaction condition with the virtual opponent, and generates an offensive recommendation using the offensive behavior model according to the defensive behavior information and the adjusted player parameters, so that the sufficiency of the player's training can be evaluated through simulated combat, and the technical effect of enabling the player to master weaknesses and strengthen technical movements is achieved.

Description

Technical Field

[0001] A basketball player training system and method, specifically a system and method that uses machine learning algorithms to simulate virtual players that target player weaknesses for targeted training. Background Technology

[0002] In today's era of rapid digital and technological development, artificial intelligence (AI), big data, and virtual reality (VR) technologies are demonstrating their unique value in many fields, especially in education.

[0003] Currently, the teaching methods used in the education field include basic classroom teaching, supplementary on-site teaching and individual guidance, special multi-grade teaching, and computerized multi-media teaching. However, most of the above teaching methods are only suitable for static teaching and not for dynamic teaching, such as physical education courses like basketball.

[0004] Traditional basketball instruction includes theoretical learning and practical training. Basketball theory has traditionally been taught to players through experience. However, basketball is a sport that demands a high level of technical practice and tactical theory. Traditional teaching methods often face limitations in terms of facilities, resources, and personalized instruction, especially in tactical understanding and practical application. Aside from actual games, it's difficult to provide players with sufficient training. Even during games, coaches can only superficially observe the players' training results. Therefore, there is a lack of a teaching model that can effectively train and provide real-time feedback.

[0005] In conclusion, it is clear that existing technologies have long suffered from the problem of difficulty in assessing whether player training programs are adequate. Therefore, it is necessary to propose improved technical methods to solve this problem. Summary of the Invention

[0006] In view of the problem that existing technologies have difficulty in assessing whether player training is sufficient, this invention discloses a system and method for targeted training based on player weaknesses using virtual players, wherein:

[0007] The system disclosed in this invention, which simulates virtual players based on player weaknesses for targeted training, includes at least: a memory for storing computer instructions; and a processor for executing the computer instructions to perform the following steps: loading multiple historical game videos, each containing a basketball and multiple players, including the target player; analyzing the historical game videos to detect the movement trajectories of the basketball and each player in each historical game video, and extracting multiple historical video features at different time points from each historical game video; obtaining player parameters for each player; determining the on-court behavior and success or failure status of each player in each historical game video based on the movement trajectories of the basketball and players or the historical video features; and performing on-court behavior based on each player. The system uses machine learning algorithms to train offensive and defensive behavior models based on the target player's movement trajectory, actions, success or failure status on the field, and player parameters. It then uses the defensive behavior model to generate weakness information based on the target player's movement trajectory, on-field actions, and success or failure status, and generates defensive behavior information accordingly. A virtual opponent is generated to execute the defensive behavior information. The system displays the real-time scene and the virtual opponent, and allows interaction with the virtual opponent in the real-time scene. Based on the interaction with the virtual opponent, if pre-set training suggestion conditions are met, the system adjusts the target player's player parameters and uses the offensive behavior model to generate offensive suggestions based on the defensive behavior information and the adjusted player parameters, displaying the offensive suggestions in the real-time scene.

[0008] The method for targeted training using virtual players based on player weaknesses disclosed in this invention includes at least the following steps: loading multiple historical game videos, each containing a basketball and multiple players, including the target player; analyzing the historical game videos to detect the movement trajectories of the basketball and each player in each historical game video, and extracting multiple historical video features at different time points from each historical game video; obtaining player parameters for each player; judging the on-court behavior and success or failure status of each player in each historical game video based on the movement trajectories of the basketball and players or the historical video features; and determining the on-court behavior based on the movement trajectory and actions of each player when performing each on-court behavior, and the success or failure status of the on-court behavior. The system uses machine learning algorithms to train offensive and defensive behavior models based on the player's status and parameters. It generates weakness information using the defensive behavior model based on the target player's movement trajectory, on-field behavior, and win / loss status, and then generates defensive behavior information accordingly. It generates virtual opponents to execute these defensive behaviors. The system displays the real-time scene and the virtual opponents, allowing interaction with them in the real-time scene. Based on the interaction with the virtual opponents, it determines if pre-set training suggestion conditions are met, adjusts the target player's parameters, and generates offensive suggestions using the offensive behavior model based on the defensive behavior information and the adjusted parameters, displaying these suggestions in the real-time scene.

[0009] The system and method disclosed in this invention are as described above. The difference between this invention and the prior art lies in the fact that, after training offensive and defensive behavior models using machine learning algorithms based on historical match footage, this invention uses the defensive behavior model to generate information on the weaknesses of the target player's on-field behavior, and generates defensive behavior information based on the weakness information. It also generates a virtual opponent to execute the defensive behavior information and provides interaction with the virtual opponent in a real-time scenario. When the training suggestion conditions are met based on the interaction with the virtual opponent, the player parameters of the target player are adjusted, and offensive suggestions are generated using the offensive behavior model based on the defensive behavior information and the adjusted player parameters. This solves the problems existing in the prior art and can achieve the technical effect of enabling players to understand their weaknesses and strengthen their technical movements. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the components of the device for targeted training of virtual players based on player weaknesses, provided by the present invention.

[0011] Figure 2 A schematic diagram of the module generated by the processor provided in this invention.

[0012] Figure 3A The flowchart illustrates the method for targeted training using virtual players based on player weaknesses, as provided by this invention.

[0013] Figure 3B This is a flowchart of the method for obtaining player parameters provided by the present invention.

[0014] Figure 3C The flowchart of the method for determining the success or failure status of a player's on-field behavior provided by the present invention.

[0015] Figure 3D The flowchart illustrates the method for generating offensive suggestions based on the on-field positions of virtual teammates and virtual opponents, as provided in this invention.

[0016] Figure 4 This is a schematic diagram of the image signal of the real-time image provided in an embodiment of the present invention.

[0017] The annotations in the attached figures are explained as follows:

[0018] 100: Device

[0019] 110: Memory

[0020] 120: Input Module

[0021] 130: Communication Interface

[0022] 140: Storage medium

[0023] 150: Output Module

[0024] 170: Processor

[0025] 190: Bus

[0026] 210: Data Acquisition Module

[0027] 220: Image Analysis Module

[0028] 230: Behavior Judgment Module

[0029] 240: Model Training Module

[0030] 250: Defense Simulation Module

[0031] 270: Scene Simulation Module

[0032] 280: Offensive Suggestion Module

[0033] 290: Interactive Display Module

[0034] 400: frames

[0035] 410: Virtual Opponent

[0036] Step 301: Load historical match footage

[0037] Step 310: Analyze historical game footage to detect the movement trajectories of the basketball and each player, and extract historical image features from the historical game footage.

[0038] Step 320: Obtain player parameters for each player

[0039] Step 321: Provide individual parameters from the input player parameters

[0040] Step 325: Based on the player's individual parameters, perform motion capture and / or posture analysis on historical match footage to obtain the player's motion parameters.

[0041] Step 330: Based on the movement trajectories of the basketball and each player, or historical video characteristics, determine the players' on-court behavior and success or failure in historical game footage.

[0042] Step 331: Based on the player's movement trajectory or historical video features, use a trained behavior recognition model and / or posture recognition model to determine the player's on-field behavior.

[0043] Step 335: Determine the success or failure of the on-court action based on whether the basketball's trajectory matches the on-court behavior.

[0044] Step 340: Based on the players' movement trajectories and actions on the field, the success or failure of their actions, and player parameters, train offensive and defensive behavior models.

[0045] Step 350: Based on the target player's movement trajectory, on-field behavior, and success / failure status, use the defensive behavior model to generate weakness information, and generate defensive behavior information based on the weakness information.

[0046] Step 360: Generate a virtual opponent to perform defensive actions.

[0047] Step 370: Display the live scene and virtual opponent, and provide interaction with the virtual opponent in the live scene.

[0048] Step 380: When the training suggestion conditions are met, adjust the target player's player parameters, and generate offensive suggestions using the offensive behavior model based on the defensive behavior information and the adjusted player parameters, and display the offensive suggestions in the real-time scene.

[0049] Step 385: Obtain the on-field positions of virtual teammates and other virtual opponents.

[0050] Step 387: Generate offensive suggestions using the offensive behavior model based on defensive behavior information, adjusted player parameters, and on-field positions. Detailed Implementation

[0051] The features and implementation methods of the present invention will be described in detail below with reference to the accompanying drawings and embodiments. The content is sufficient to enable any person skilled in the art to easily and fully understand the technical means used by the present invention to solve the technical problem and to implement it accordingly, thereby achieving the effects that the present invention can achieve.

[0052] This invention can analyze the weaknesses of specific players in basketball games from historical game footage and generate virtual players whose behaviors target the analyzed weaknesses. Simultaneously, it allows the specific player to interact with the generated virtual player using real-world technology to train against these weaknesses. The real-world technology provided by this invention includes, but is not limited to, Virtual Reality (VR), Mixed Reality (MR), and Extended Reality (XR).

[0053] The apparatus for implementing this invention can be a computing device, which includes, but is not limited to, one or more processors, one or more memories, and buses connecting different hardware components (including memories and processors). Through the included hardware components, the computing device can load and execute an operating system, allowing the operating system to run on the computing device, and can also execute software or programs. The computing device also includes a housing, within which the aforementioned hardware components are disposed.

[0054] The bus of the computing device proposed in this invention may include one or more types, such as a data bus, address bus, control bus, expansion bus, and / or local bus. The bus of the computing device includes, but is not limited to, Industry Standard Architecture (ISA) bus, Peripheral Component Interconnect (PCI) bus, Video Electronics Standards Association (VESA) local bus, Universal Serial Bus (USB), and PCI Express (PCI-E / PCIe) bus.

[0055] The processor of the computing device proposed in this invention is coupled to a bus. The processor includes a set of registers or a register space, which may be entirely located on the processor's processing chip, or wholly or partially located outside the processing chip and coupled to the processing chip via dedicated electrical connections and / or via a bus. The processor may be a central processing unit, a microprocessor, or any suitable processing element. If the computing device is a multiprocessor device, that is, the computing device contains multiple processors, then the processors contained in the computing device are identical or similar and are coupled and communicate via a bus. In some embodiments, the processor can interpret a computer instruction or a series of multiple computer instructions to perform specific operations or calculations, such as mathematical operations, logical operations, data comparison, copying / moving data, etc., thereby driving other hardware components in the computing device or running an operating system or executing various programs and / or modules. Computer instructions can be assembly language instructions, instruction set architecture instructions, machine instructions, machine-dependent instructions, microinstructions, firmware instructions, or source code or object code written in any combination of one or more programming languages. Computer instructions can be executed entirely on a single computing device, partially on a single computing device, or partially on one computing device and partially on another connected computing device. The aforementioned programming languages ​​include object-oriented programming languages ​​such as Common Lisp, Python, C++, Objective-C, Smalltalk, Delphi, Java, Swift, C#, Perl, Ruby, etc., as well as conventional procedural programming languages ​​such as C or other similar programming languages.

[0056] Computing devices typically include one or more chipsets. The processor of the computing device can be coupled to the chipset or electrically connected to the chipset via a bus. A chipset consists of one or more integrated circuits (ICs), including a memory controller and peripheral input / output (I / O) controllers, etc. That is, the memory controller and I / O controllers can be contained within a single IC or implemented using two or more ICs. Chipsets typically provide I / O and memory management functions, as well as multiple general-purpose and / or special-purpose registers, timers, etc., which can be accessed or used by one or more processors coupled to or electrically connected to the chipset. In some embodiments, the chipset may also be part of the processor.

[0057] The processor of a computing device can also access data in the memory and mass storage areas installed on the computing device through the memory controller. The aforementioned memory includes any type of volatile memory and / or non-volatile memory (NVRAM), such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Read-Only Memory (ROM), Flash memory, etc. The aforementioned mass storage areas can contain any type of storage device or storage medium, such as hard disks, optical discs, flash drives, memory cards, solid-state drives (SSDs), or any other storage device. In other words, the memory controller can access data in static random access memory, dynamic random access memory, flash memory, hard disk drives, and solid-state drives.

[0058] The processor of a computing device can also connect and communicate with peripheral devices or interfaces such as peripheral output devices, peripheral input devices, communication interfaces, and various data or signal receiving devices via a peripheral input / output controller and a peripheral input / output bus. Peripheral input devices can be any type of input device, such as a keyboard, mouse, trackball, touchpad, joystick, etc., while peripheral output devices can be any type of output device, such as a monitor, printer, etc. Peripheral input devices and peripheral output devices can also be the same device, such as a touchscreen. Communication interfaces can include wireless communication interfaces and / or wired communication interfaces. Wireless communication interfaces can include interfaces supporting wireless local area networks (such as Wi-Fi, Zigbee, etc.), Bluetooth, infrared, near-field communication (NFC), 3G / 4G / 5G mobile communication networks (cellular networks), or other wireless data transmission protocols. Wired communication interfaces can be Ethernet devices, DSL modems, cable modems, asynchronous transfer mode (ATM) devices, or fiber optic communication interfaces and / or components, etc. The data or signal receiving device may include a GPS receiver or a physiological signal receiver, the physiological signals received by the physiological signal receiver including, but not limited to, heartbeat, blood oxygenation, etc. The processor can periodically poll various peripheral devices and interfaces, enabling the computing device to input and output data through various peripheral devices and interfaces, and also to communicate with another computing device having the hardware components described above.

[0059] The following is a preliminary step. Figure 1 The present invention uses schematic diagrams of components to illustrate the device for targeted training using virtual players based on player weaknesses. For example... Figure 1 As shown, the device 100 of the present invention includes a memory 110, an input module 120, a communication interface 130, a storage medium 140, an output module 150, a processor 170, and a bus 190. The memory 110, input module 120, communication interface 130, storage medium 140, output module 150, and processor 170 are interconnected via the bus 190.

[0060] The memory 110 can store one or more sets of computer instructions.

[0061] The input module 120 can provide input data through the peripheral output devices of the device 100. For example, the input module 120 can provide the storage path for selecting or inputting historical match videos and / or player headshots, and set individual player parameters through peripheral output devices such as a keyboard and / or mouse.

[0062] The communication interface 130 can be connected to an external network storage device or server (not shown in the figure) and request and download data from the connected network device. For example, the communication interface 130 can be connected to a network device to download data such as historical match footage, offensive tactical information, and defensive tactical information. The historical match footage can be transmitted to the device 100 by the network device in the form of a file or a stream. There are no particular limitations in this invention.

[0063] Storage medium 140 can store data received by communication interface 130, and can also store data required by processor 170 (i.e. provided to processor 170), such as historical game images, offensive tactical information, defensive tactical information, etc.; storage medium 140 can also store data generated by processor 170, such as offensive suggestions, data of virtual objects required for display technology, etc.

[0064] The output module 150 can output the data generated by the processor 170 through the peripheral output device of the device 100. For example, the output module 150 can output a real-time scene and a virtual player using a display or touch screen, allowing the device 100 to provide interaction with the virtual player in the real-time scene. The output module 150 can also annotate attack suggestions in the real-time scene. The real-time scene provided by this invention may vary depending on the real-time technology. For example, when the real-time technology is virtual reality, the real-time scene is a virtual scene generated by the real-time technology, while when the real-time technology is mixed reality, the real-time scene can be an actual scene in the real world or a virtual scene.

[0065] Processor 170 can be like Figure 2 The schematic diagram of the processor provided by this invention shows that it includes a data acquisition module 210, an image analysis module 220, a behavior judgment module 230, a model training module 240, a defense simulation module 250, a scene simulation module 270, an attack suggestion module 280, and an interactive display module 290. In some embodiments, the processor 170 can execute computer instructions stored in the memory 110, and can generate an interactive display module after executing the computer instructions. Figure 2 The various modules within; in another embodiment, Figure 2 The modules within can be generated from one or more circuits and / or complete or partial chips and other hardware components; that is, the processor 170 includes components... Figure 2 The hardware components of each module in the processor 170, that is, each module included in the processor 170 can be a software module or a hardware module, and there are no particular limitations in this invention.

[0066] The data acquisition module 210 is responsible for loading multiple game videos, which include historical game videos recorded during past games. Each game video loaded by the data acquisition module 210 contains a basketball and players, and the players in the game video include one or more offensive players or one or more defensive players. Generally speaking, different game videos loaded by the data acquisition module 210 may contain the same one or more players, or may contain the same team.

[0067] For example, the data acquisition module 210 can read the match video from the storage medium 140 according to the storage path provided by the input module 120, or it can download the match video to the network device via the communication interface 130 according to the storage path provided by the input module 120; the data acquisition module 210 can also directly read the match video from a specific directory of the storage medium 140, or it can download the match video from a specific directory of the network device via the communication interface 130. However, the method by which the data acquisition module 210 loads the match video is not limited to the above.

[0068] The data acquisition module 210 is also responsible for acquiring the player's individual parameters. In some embodiments, the data acquisition module 210 can also acquire the player's motion parameters. The individual parameters provided by this invention are items related to the length of the human body or limbs, including but not limited to the player's height, arm length, leg length, etc.; the motion parameters provided by this invention are items related to the player's movement, including but not limited to the player's jump height, stride length, movement speed, etc. In this invention, the player's parameters include both the player's individual parameters and motion parameters.

[0069] The data acquisition module 210 can read player parameters from the storage module 140 based on player identification data, or download player parameters from an external storage device or data server via the communication interface 130 based on player identification data, or receive player parameters via the communication interface 130, or input player parameters via the input module 120. However, the methods by which the data acquisition module 210 acquires player parameters are not limited to the above. The player identification data can be a serial number generated by the system for the player, the player's name, or the player's image features, etc. Any data that can be used to identify a player can be used as player identification data; this invention has no particular limitations.

[0070] The image analysis module 220 is responsible for analyzing the game images loaded by the data acquisition module 210 to generate image analysis information corresponding to each game image. Each image analysis information generated by the image analysis module 220 may include the on-court positions and movement trajectories of the basketball and each player detected by the image analysis module 220 for a corresponding game image, and multiple image features extracted from the corresponding game image at different time points. In this invention, the image features extracted from historical game images are called historical image features. The movement trajectories of the basketball and each player detected by the image analysis module 220 may include not only the time points corresponding to the on-court positions, but may also include speed and acceleration, but this invention is not limited thereto.

[0071] The image analysis module 220 can extract features from each frame or specific frame in the game image to extract image features from frames at different time points in each game image. The specific frames mentioned above can be frames in the game image at fixed intervals or at fixed time intervals (such as frames every five frames or frames every 0.2 seconds), or frames with a similarity to the previous frame that is less than a certain value, or frames where events such as cutting, shooting, passing, turnover, interception, or blocking occur. However, the image analysis module 220 is not limited to the above methods of selecting frames for feature extraction. For example, the image analysis module 220 can use pre-trained object detection models such as YOLO and Faster R-CNN to detect objects such as players and balls in the frame, in order to obtain the bounding boxes of each object in the frame; the image analysis module 220 can also use pre-trained residual neural network (ResNet) models, visual geometry group (VGG) models, and other convolutional neural network (CNN) models to extract spatial features from the frames in the game image, in order to extract the frame spatial feature data of the frame and the object spatial feature data of each object in the frame, but the present invention is not limited thereto; the image analysis module 220 can also use multi-object tracking (Multi-Object Tracking) methods such as SORT (Simple Online and Realtime Tracking), DeepSORT (Simple Online and Realtime Tracking with a Deep Association Metric), and Kalman Filter. The Tracking over Time (MOT) algorithm tracks detected objects in subsequent frames to identify the same objects again. It can also process time-series data by arranging the spatial feature data of each object in each frame according to the chronological order of the frames. This time-series data is then processed using either a Recurrent Neural Network (RNN) or a Long Short-Term Memory (LSTM) model to generate temporal feature data for each object in the game footage. The spatial feature data of an object (i.e., a basketball or a player) represents its position, while the temporal feature data represents its trajectory and velocity.

[0072] The image analysis module 220 can also calculate the player's motion parameters. For example, the image analysis module 220 can perform motion capture and / or posture analysis on each player from the game video, and calculate the player's stride length, jump height, and movement distance in the game video according to the individual parameters such as the player's height, arm length, and leg length. It can also calculate the player's movement speed based on the movement distance and movement time.

[0073] The behavior judgment module 230 is responsible for judging the on-court behavior of each player based on the basketball and player's on-court position and movement trajectory or image features generated by the image analysis module 220, and can obtain the time when each player begins to perform the judged on-court behavior. Generally speaking, the on-court behavior of offensive players includes shooting, passing, cutting, blocking, and positioning, while the on-court behavior of defensive players includes marking, intercepting, blocking, and positioning, but the present invention is not limited thereto.

[0074] The behavior judgment module 230 can use a trained behavior recognition model and / or posture recognition model to identify the player's on-field behavior represented by the image features generated by the image analysis module 220. For example, the behavior judgment module 230 can use a posture estimation model such as OpenPose to determine the position (e.g., coordinates) of key points (including but not limited to head, shoulders, elbows, wrists, hips, knees, ankles, etc.) of various parts of the player's body, and can use the position changes of each key point over time as input to use the behavior recognition model to identify the player's on-field behavior. However, the representation of key points and their positions proposed in this invention is not limited to the above.

[0075] The behavior judgment module 230 can train the behavior recognition model based on a large number of player images with known actions. For example, the behavior judgment module 230 provides a large number of player images and the actions of the players in each player image to the behavior recognition model and / or posture recognition model, so that after the behavior recognition model / posture recognition model determines the position of the key points of each part of the player's body in each player image, it can be trained based on the position of the key points of each part of the player's body in each player image and the known actions.

[0076] The behavior judgment module 230 can also judge the on-court behavior of the offensive player based on the basketball and the offensive player's position and movement trajectory and the position of the basket. For example, when the basketball's trajectory moves from one offensive player with the ball to another, the behavior judgment module 230 can determine that the offensive player with the ball's on-court behavior is passing; when the basketball's trajectory moves from one offensive player with the ball to the basket, and there are no other offensive players within a certain distance of the basket, the behavior judgment module 230 can determine that the offensive player with the ball's on-court behavior is shooting; when the basketball's trajectory is the same as that of the offensive player with the ball, and there is a defensive player within a certain distance of the offensive player with the ball, the behavior judgment module 230 can determine that the offensive player with the ball's on-court behavior is cutting or penetrating; when the basketball's trajectory is the same as that of the offensive player with the ball, and there are other offensive players without the ball within a certain distance of the offensive player with the ball, and there is a defensive player on a side of the offensive player without the ball that is different from the offensive player with the ball, the behavior judgment module 230 can determine that the offensive player without the ball's on-court behavior is blocking. However, the behavior judgment module 230's methods for judging the offensive player's on-court behavior are not limited to the above.

[0077] The behavior judgment module 230 is responsible for judging the success or failure of each player's on-court behavior based on the movement trajectory or historical image characteristics of the basketball and each player. For example, the behavior judgment module 230 can judge the success or failure of an on-court behavior based on whether the movement trajectory of the basketball matches the judged on-court behavior. For instance, the behavior judgment module 230 can judge whether an offensive player's pass is received by another offensive player, that is, whether the movement trajectory of the basketball matches that of another offensive player. If yes, the on-court behavior of passing is successful; otherwise, the pass is unsuccessful. The behavior judgment module 230 can also judge whether the movement trajectory of the basketball passes through the position of the basket. If yes, the on-court behavior of shooting is successful; otherwise, the shot is unsuccessful. The behavior judgment module 230 can also judge whether the movement trajectory of the basketball continuously matches that of the offensive player with the ball when the offensive player with the ball is cutting or breaking through. If yes, the on-court behavior of cutting or breaking through is successful; otherwise, the cutting or breaking through is unsuccessful. However, the behavior judgment module 230's method of judging whether an on-court behavior is successful or unsuccessful is not limited to the above.

[0078] The model training module 240 is responsible for training offensive and defensive behavior models using machine learning algorithms based on the movement trajectory and actions of each player during on-court actions, the success or failure status of each action, and the player parameters of each player performing each action. For example, the model training module 240 can use a first machine learning algorithm to train an offensive behavior model based on the movement trajectory, on-court actions, success or failure status, and player parameters of each offensive player during a given offensive action; it is also responsible for using a second machine learning algorithm to train a defensive behavior model based on the movement trajectory, on-court actions, success or failure status, and player parameters of each defensive player during a given defensive action. The aforementioned first and second machine learning algorithms include, but are not limited to, long short-term memory models and Transformer models. More specifically, the model training module 240 can determine the start time of each offensive or defensive action in each game image based on the basketball and the on-court positions and movement trajectories of each player generated by the image analysis module 220 or the historical image features generated by the image analysis module 220. That is, it can determine the start time of each offensive or defensive transition (such as determining the transition when the ball-handling player changes teams, but this invention is not limited to this). It can also provide the movement trajectory, on-court behavior, success or failure status, and player parameters of the offensive player during the offensive process after each offensive or defensive transition to the first machine learning algorithm to train the offensive behavior model, and provide the movement trajectory, on-court behavior, success or failure status, and player parameters of the defensive player during the defensive process after each offensive or defensive transition to the second machine learning algorithm to train the defensive behavior model. However, the way the model training module 240 trains the offensive behavior model and the defensive behavior model is not limited to the above.

[0079] The defensive simulation module 250 is responsible for generating weakness information based on the target player's movement trajectory generated by the image analysis module 220, the on-field behavior and success / failure status generated by the behavior judgment module 230, and the defensive behavior model trained by the model training module 240. It is also responsible for generating defensive behavior information based on this weakness information. For example, the defensive simulation module 250 can analyze the target player's on-field behavior when the number or percentage of failed attempts during an attack reaches a certain value using the defensive behavior model. It can also generate corresponding defensive behavior information based on the object-time characteristic data of on-field behavior when the number or percentage of failed attempts exceeds a certain value. For instance, when the target player makes a certain number of mistakes due to defensive pressure, it can generate defensive behavior information that corresponds to increased defensive pressure.

[0080] The scene simulation module 270 is responsible for generating virtual opponents who execute the defensive behavior information generated by the defense simulation module 250. In some embodiments, if the scene simulation module 270 simulates a scene in virtual reality or mixed reality, it can also simulate a basketball court to generate a virtual real-time scene.

[0081] The scene simulation module 270 is also responsible for obtaining the interaction status between the user (usually the target player) and the virtual opponent. The interaction status provided by this invention includes, but is not limited to, the distance, relative position, actions, and movement routes between the user and the virtual opponent.

[0082] The attack suggestion module 280 is responsible for determining whether the training suggestion conditions are met based on the interaction between the user and the virtual opponent. These training suggestion conditions are predefined and include, but are not limited to, the virtual opponent not being successfully breached after a certain period of time. For example, the attack suggestion module 280 can determine whether the user has successfully breached the virtual opponent based on the relative position and movement route between the user and the virtual opponent. If the user has not breached the virtual opponent after a certain period of time, the attack suggestion module 280 can determine that the training suggestion conditions are met.

[0083] The offensive suggestion module 280 is responsible for adjusting the player parameters of the target player obtained by the data acquisition module 210 (and the data analysis module 220). More specifically, the offensive suggestion module 280 can add corresponding training values ​​to one or more items of the target player's player parameters. The training values ​​for different items may be different. For example, the training value for height is 1 cm, the training value for arm length / leg length is 0.5 cm, the training value for jump height is 2 cm, the training value for stride length is 1 cm, and the training value for movement speed is 5 meters per second, etc., but the present invention is not limited thereto.

[0084] The offensive suggestion module 280 is also responsible for generating offensive suggestions based on the defensive behavior information generated by the defensive simulation module 250 and the adjusted player parameters, using the offensive behavior model trained by the model training module 240.

[0085] Generally speaking, the offensive suggestion module 280 can prioritize generating offensive suggestions for individual offenses, such as suggestions that include dribbling methods and body movements for individual breakthroughs or shots. If the offensive behavior model determines that there is no effective individual offensive plan, the offensive suggestion module 280 can also obtain the on-court positions of one or more virtual teammates and one or more other virtual opponents, and can use the offensive behavior model to generate offensive suggestions for team offenses based on the defensive behavior information generated by the defensive simulation module 250, the adjusted player parameters, and the obtained on-court positions, such as generating offensive suggestions that include movement trajectories that attract defensive players or disrupt defensive strategies and pass the ball to open teammates.

[0086] The interactive display module 290 is responsible for generating display signals of the virtual opponent generated by the scene simulation module 270 in the real-time scene, and providing users with interaction with the virtual opponent in the real-time scene through the interactive information generated by the scene simulation module 270. That is, it generates display signals of the behavior of the virtual opponent in the real-time scene based on the interactive information.

[0087] The interactive display module 290 is also responsible for generating display signals that show attack suggestions in real-time scenarios.

[0088] The interactive display module 290 can transmit the generated display signal to the display device through the communication interface 130, so that the display device can display a virtual opponent in the real-time scene according to the display signal and provide the user with the ability to interact with the virtual opponent in the real-time scene.

[0089] The operating system and method of the present invention will then be explained using an embodiment, and please refer to [reference needed]. Figure 3A The present invention provides a flowchart of a method for targeted training using virtual players based on player weaknesses. In this embodiment, it is assumed that device 100 is a server providing virtual reality, mixed reality, or extended reality, but the present invention is not limited thereto.

[0090] When basketball coaches, assistants, or players wish to train players on their weaknesses, if the team members choose to use this invention, they can first connect to device 100 using a personal computer. Through the personal computer, they can input the storage path and filename of historical game footage on the user interface provided by the data acquisition module 210 of device 100, and then transmit the input storage path and filename of the historical game footage back to device 100, allowing device 100 to load the historical game footage (step 301). Assuming that in this embodiment, the historical game footage is stored on an external file server, after the processor 170 of device 100 executes the computer instructions stored in the memory 110 of device 100, it can generate... Figure 2 The executable modules allow the data acquisition module 210 of device 100 to connect to an external file server via the communication interface 130 of device 100, and download the historical match images from the file server according to the storage path and file name of the historical match images received by the communication interface 130.

[0091] After the data acquisition module 210 of device 100 loads historical game images (step 301), the image analysis module 220 of device 100 can analyze each loaded historical game image to detect the on-court position and movement trajectory of the basketball and each player in each historical game image, and extract historical image features at different time points from each historical game image (step 310).

[0092] Similarly, after the data acquisition module 210 of device 100 loads the historical match footage (step 301), the data acquisition module 210 (and the image analysis module 220 of device 100) can also obtain the player parameters of each player in the historical match footage (step 320). In this embodiment, it is assumed that... Figure 3B As shown in the process, the data acquisition module 210 can read the individual parameters of each player from the storage medium 140 based on the image features of each player's face in each historical match video acquired by the image analysis module 220. When the individual parameters of a specific player are not in the storage medium 140, the individual parameters of a specific player can be obtained by the team personnel through the input module 120 or the communication interface 130 (step 321). The image analysis module 220 can also perform motion capture and / or posture analysis on each player in the historical match video and calculate the player's motion parameters based on the player's individual parameters (step 325) to generate player parameters that include individual parameters and motion parameters.

[0093] After the image analysis module 220 of device 100 detects the on-court positions and movement trajectories of the basketball and each player in each historical game image, and extracts historical image features at different time points from each historical game image (step 310), the behavior judgment module 230 of device 100 can judge each player's on-court behavior and corresponding success or failure status in each historical game image based on the on-court positions and movement trajectories of the basketball and each player in each historical game image or the historical image features (step 330). In this embodiment, assuming as follows Figure 3C As shown in the process, the behavior judgment module 230 can use the trained behavior recognition model or posture recognition model to judge the on-court behavior of each player in each historical game video based on the movement trajectory of each player in each historical game video or the features of the historical video (step 331), and can judge the success or failure of the on-court behavior of the player based on whether the movement trajectory of the basketball matches the on-court behavior of each player (step 335).

[0094] After the data acquisition module 210 (and the image analysis module 220 of the device 100) acquires the player parameters of each player in the historical match video (step 320) and the behavior judgment module 230 of the device 100 judges each player's on-field behavior and corresponding success or failure status in the historical match video (step 330), the model training module 340 of the device 100 can train offensive behavior models and defensive behavior models based on the movement trajectory and actions of each player's on-field behavior, the success or failure status of the on-field behavior, and the player parameters of each player (step 340).

[0095] After the model training module 340 of device 100 completes the training of the offensive behavior model and the defensive behavior model, the defensive simulation module 250 of device 100 can use the defensive behavior model trained by the model training module 340 to generate weakness information relative to the target player based on the target player's movement trajectory, on-court behavior, and the success or failure status of the on-court behavior, and can generate defensive behavior information based on the generated weakness information (step 350). In this embodiment, it is assumed that the defensive simulation module 250 can generate different defensive behaviors according to the weaknesses represented by different weakness information. For example, it can generate defensive behavior information that implements individual defensive plans (such as focusing on defending the left / right side, strengthening inside / outside defense, etc.) or team defensive strategies (such as zone defense, man-to-man defense, etc.) that the target player is prone to dribbling errors, passing errors, or missed shots.

[0096] After the defensive simulation module 250 of device 100 generates defensive behavior information, the scene simulation module 280 of device 100 can generate a virtual opponent that performs the defensive behavior information (step 360). The interactive display module 290 of device 100 can generate an image signal of the virtual opponent and transmit the generated image signal to the real-time display device used by the team members through the communication interface 130 of device 100. This allows the real-time display device to display the image signal and show the virtual opponent in the real-time scene, providing the target player with the opportunity to interact with the virtual opponent in the real-time scene (step 370). In this embodiment, it is assumed that the interactive display module 290 can generate information including... Figure 4 The image signal shown in frame 400 contains a virtual opponent 410 and other virtual players.

[0097] After the scene simulation module 280 of device 100 generates a virtual opponent that performs defensive behavior information (step 360), and the interactive display module 290 of device 100 generates the image signal of the virtual opponent, the offensive suggestion module 270 of device 100 can determine whether the training suggestion conditions are met. If not, the offensive suggestion module 270 can not generate an offensive suggestion. If yes, that is, the training suggestion conditions are met, the offensive suggestion module 270 can adjust the player parameters of the target player obtained by the data acquisition module 210 (and the image analysis module 220 of device 100) of device 100, and can generate an offensive suggestion based on the adjusted player parameters and the defensive behavior information generated by the defensive simulation module 250 of device 100 using the offensive behavior model trained by the model training module 240 of device 100. This causes the interactive display module 290 to generate the image signal of the virtual opponent that performs the offensive suggestion generated by the offensive suggestion module 270 and transmit the generated image signal to the real display device through the communication interface 130 of device 100, so that the real display device can display the offensive suggestion in the real scene (step 380). In this embodiment, it is assumed that the offensive suggestion module 280 can first attempt to generate offensive suggestions that allow the target player to directly attack, such as dribbling methods and body movements for individual breakthroughs or jump shots. The offensive suggestion module 280 can also, when the offensive behavior model determines that there are no effective offensive individual offensive plans, such as... Figure 3D As shown in the process, the on-court positions of virtual teammates and other virtual opponents in the real-time scene basketball court are further obtained (step 385). Based on the defensive behavior information generated by the defensive simulation module 250, the adjusted player parameters and the obtained on-court positions, the offensive behavior model can be used to generate offensive suggestions for the target player to assist the virtual teammate (step 387). For example, a team offensive plan is generated that includes dribbling methods and movement trajectories that attract defensive players or disrupt defensive strategies and passes the ball to teammates with scoring opportunities.

[0098] Thus, through this invention, players can train to target their weaknesses using virtual reality or mixed reality, and can even learn offensive strategies from other players whose individual or athletic parameters are slightly better than their own, thereby improving their skills.

[0099] In summary, the difference between this invention and existing technologies lies in its ability to train offensive and defensive behavior models using machine learning algorithms based on historical match footage. This training then uses the defensive behavior model to generate information on the target player's weaknesses on the field, and generates defensive behavior information based on this weakness information. Furthermore, it generates a virtual opponent to execute the defensive behavior information and provides interaction with the virtual opponent in a real-time scenario. When the training suggestion conditions are met based on the interaction with the virtual opponent, the target player's parameters are adjusted. Finally, based on the defensive behavior information and the adjusted player parameters, the offensive behavior model generates offensive suggestions. This approach addresses the problem in existing technologies of difficulty in assessing the adequacy of player training, thereby enabling players to identify weaknesses and strengthen their technical movements.

[0100] Furthermore, the method of the present invention for targeted training based on player weaknesses by simulating virtual players can be implemented in hardware, software, or a combination of hardware and software. It can also be implemented in a centralized manner in a computer system or in a decentralized manner with different components distributed among several interconnected computer systems.

[0101] While the embodiments disclosed in this invention are as described above, the content is not intended to directly limit the scope of patent protection for this invention. Any modifications made by those skilled in the art to the form and details of the implementation of this invention without departing from the spirit and scope disclosed herein shall fall within the scope of patent protection for this invention. The scope of patent protection for this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A method for targeted training using virtual players based on player weaknesses, applied to a device, the method comprising at least the following steps: Load multiple historical game videos, each of which contains a basketball and multiple players, including the target player; The multiple historical game images are analyzed to detect the movement trajectories of the basketball and each player in each historical game image, and multiple historical image features at different time points are extracted from each historical game image. Obtain the player parameters for each of the aforementioned players; Based on the movement trajectory of the basketball and the multiple players or the features of the multiple historical images, determine the on-court behavior of each player in each of the historical game images and the success or failure status of the on-court behavior; Based on the movement trajectory and actions of each player when performing each of the on-court actions, the success or failure status of the on-court actions, and the player parameters of each player, an offensive behavior model and a defensive behavior model are trained using machine learning algorithms. Based on the target player's multiple movement trajectories, multiple on-field behaviors, and multiple success and failure states, the defensive behavior model is used to generate weakness information, and defensive behavior information is generated based on the weakness information; Generate a virtual opponent that performs the aforementioned defensive actions; Displays a real-time scene and the virtual opponent, and provides the ability to interact with the virtual opponent in the real-time scene; and When the pre-set training suggestion conditions are met based on the interaction with the virtual opponent, the player parameters of the target player are adjusted, and offensive suggestions are generated using the offensive behavior model based on the defensive behavior information and the adjusted player parameters, and the offensive suggestions are displayed in the real-time scene. The step of adjusting the player parameters of the target player is to increase the training value corresponding to at least one of the player parameters of the target player. The player parameters of the target player include at least one of the following: height, arm length, leg length, jump height, stride length, and movement speed.

2. The method for targeted training using virtual players based on player weaknesses as described in claim 1, wherein the step of determining the on-court behavior of each player in each game video and the success or failure status of the on-court behavior based on the movement trajectory of the basketball and the multiple players or the multiple historical video features is as follows: determining the multiple on-court behaviors of each player using a trained behavior recognition model and / or posture recognition model based on the movement trajectory of the multiple players or the multiple historical video features, and determining the success or failure status of the on-court behavior based on whether the movement trajectory of the basketball matches the multiple on-court behaviors.

3. The method for targeted training based on player weaknesses using virtual players as described in claim 1, wherein the step of obtaining player parameters for each player further includes providing individual parameters from the player parameters as input, and performing motion capture and / or posture analysis on each player based on historical match footage to obtain motion parameters from the player parameters.

4. The method for targeted training using virtual players based on player weaknesses as described in claim 1, wherein the step of generating the offensive suggestion using the offensive behavior model based on the defensive behavior information and the adjusted player parameters further includes the step of obtaining the on-field positions of at least one virtual teammate and at least one other virtual opponent, and generating the offensive suggestion using the offensive behavior model based on the defensive behavior information, the adjusted player parameters, and multiple on-field positions.

5. A system for targeted training using virtual players based on player weaknesses, applied to a device, the system comprising at least: Memory, used to store at least one computer instruction; and A processor for executing the at least one computer instruction to perform the following steps: Load multiple historical game videos, each of which contains a basketball and multiple players, including the target player; The multiple historical game images are analyzed to detect the movement trajectories of the basketball and each player in each historical game image, and multiple historical image features at different time points are extracted from each historical game image. Obtain the player parameters for each of the aforementioned players; Based on the movement trajectory of the basketball and the multiple players or the features of the multiple historical images, determine the on-court behavior of each player in each of the historical game images and the success or failure status of the on-court behavior; Based on the movement trajectory and actions of each player when performing the on-court actions, the success or failure status of the on-court actions, and the player parameters of each player, an offensive behavior model and a defensive behavior model are trained using machine learning algorithms. Based on the target player's multiple movement trajectories, multiple on-field behaviors, and multiple success and failure states, the defensive behavior model is used to generate weakness information, and defensive behavior information is generated based on the weakness information; Generate a virtual opponent that performs the aforementioned defensive actions; Displays a real-time scene and the virtual opponent, and provides the ability to interact with the virtual opponent in the real-time scene; and When the pre-set training suggestion conditions are met based on the interaction with the virtual opponent, the player parameters of the target player are adjusted, and offensive suggestions are generated using the offensive behavior model based on the defensive behavior information and the adjusted player parameters, and the offensive suggestions are displayed in the real-time scene. The processor is described as adjusting the player parameters of the target player by increasing the training value corresponding to at least one of the player parameters of the target player. The player parameters of the target player include at least one of the following: height, arm length, leg length, jump height, stride length, and movement speed.

6. The system for targeted training based on player weaknesses using virtual player simulation as described in claim 5, wherein the processor performs the steps of using a trained behavior recognition model and / or posture recognition model to determine the plurality of on-court behaviors of each player, and determining the success or failure status of the on-court behaviors based on whether the trajectory of the basketball matches the plurality of on-court behaviors, thereby realizing the step of determining the on-court behaviors of each player in each of the game images and the success or failure status of the on-court behaviors based on the trajectory of the basketball and the plurality of players or the plurality of historical image features.

7. The system for targeted training of virtual players based on player weaknesses as described in claim 5, wherein the processor performs the step of providing individual parameters from the player parameters of each player as input, and performing motion capture and / or posture analysis on each player from historical match footage based on the individual parameters of each player to obtain the motion parameters from the player parameters of each player, thereby achieving the step of obtaining the player parameters of each player.

8. The system for targeted training of virtual players based on player weaknesses as described in claim 5, wherein the processor is further configured to perform the steps of obtaining the on-field positions of at least one virtual teammate and at least one other virtual opponent, and generating the offensive suggestion using the offensive behavior model based on the defensive behavior information, the adjusted player parameters, and the plurality of on-field positions.

Images