Volleyball match real-time data analysis system

Abstract

The invention relates to the technical field of sports technique and tactics data analysis, and discloses a volleyball match real-time data analysis system which comprises a sensing module used for obtaining real-time state vectors of players and volleyballs; the self-adaptive potential field nuclear module is used for dynamically generating personalized model parameters according to real-time states and historical data of players; the potential field generation module is used for generating attack, defense and corresponding potential fields representing attack threats, defense coverage and organization capabilities based on the state vectors and the model parameters; the potential field coupling module is used for coupling a plurality of independent potential fields into a unified net opportunity potential field; and the space-time analysis module is used for calculating a tactical momentum field representing a situation evolution trend based on the time sequence of the net opportunity potential field. According to the method, the self-adaptive dynamic situation field model is constructed, so that the physical state of the player can be converted into continuous quantitative evaluation on the space control force and tactical opportunities of the competition field in real time, and data support is provided for on-site tactical decision.

Description

Technical Field

[0001] This invention relates to the field of sports tactics and data analysis technology, specifically a real-time data analysis system for volleyball matches. Background Technology

[0002] As a team-based competitive sport, volleyball involves a series of high-speed, dynamic, and interconnected offensive and defensive transitions, which contain complex and ever-changing tactical games. In the current technology, the development of data analysis for volleyball matches mainly revolves around processing data from match videos or data collected through wearable sensors to achieve event detection and data statistics for various technical actions during the match. This analysis is conducted after the match, with the goal of providing a series of discrete, event-count-based objective statistical data for technical review by coaching staff and athletes, athlete performance evaluation, and the formulation of long-term training plans. Examples of such data include spiking success rate, blocking points, and first pass accuracy.

[0003] However, while such event statistics-based technologies have their value in providing post-match summary reports, they essentially break down a continuous, flowing game into a series of relatively isolated technical action events for analysis. This approach has technical limitations in revealing the deeper dynamic tactical confrontations of the game. Existing technologies can record and quantify action events that have already occurred and can be clearly defined, but they are difficult to quantify and describe the dynamic confrontation process of the offensive and defensive intentions of both sides on the field in real time and continuously. They also cannot reflect the continuous impact of players' off-the-ball behavior (such as running and positioning) on ​​the tactical situation.

[0004] Specifically, existing technologies lack a quantitative evaluation model that can transform players' real-time position, speed, body posture, and other basic physical state information into control of the playing space and tactical opportunities. Tactical advantages on the field are not generated only at the moment of ball contact, but are the result of the continuous movement and positioning by both sides throughout the entire round to compete for and control the playing space. For example, the formation of an offensive opportunity depends not only on the quality of the setter's pass, but also on the timing of the offensive player's run-up, running route, and the defensive player's positioning gaps. Existing technologies cannot provide an objective data field to continuously and in real-time describe these fleeting tactical opportunity windows, advantageous areas or weak points formed on both the offensive and defensive ends, and the dynamic evolution trend of these situations, which are formed by the combined actions of multiple players.

[0005] Furthermore, existing analytical models often fail to take into account the individual technical characteristics, tactical habits, and fluctuations in performance caused by factors such as physiological fatigue during the game. The models use a set of universal, static parameters to evaluate all athletes, which ignores the differences in ability between athletes and the fluctuations in the performance of the same athlete at different times of the game. Therefore, the accuracy and personalization of existing analytical results are limited, and they cannot truly reflect the real tactical threat or defensive ability posed by a specific player at a specific moment. Due to the lack of in-depth quantitative analysis and dynamic trend prediction of the real-time situation, on-the-spot tactical decisions during the game still largely rely on the judgments of coaches and athletes based on long-term accumulated, relatively subjective intuitive experience, lacking objective, continuous, and forward-looking data support. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a real-time data analysis system for volleyball matches, which solves the problem that existing technologies can only perform post-event statistics and cannot quantitatively assess real-time tactical situations.

[0007] To achieve the above objectives, the present invention provides a real-time data analysis system for volleyball matches, comprising: a perception module, an adaptive potential field kernel module, a potential field generation module, a potential field coupling module, a spatiotemporal analysis module, and a decision support module.

[0008] By adopting the above technical solutions, this system transforms the discrete, multi-source physical state data in the game into a quantitative analysis system that can continuously and in real-time reflect the distribution of invisible tactical opportunities and threats on the field through a series of mathematical modeling and calculations. The core lies in the construction of a dynamic model that maps players' behavior, position, and ability into control and influence over the field space. This solves the technical problem that existing technologies can only perform event statistics but cannot conduct in-depth quantitative analysis of tactical intentions and game situations.

[0009] In one specific embodiment, the player state vector includes the player's three-dimensional spatial position, velocity vector, and body posture category, and the volleyball state vector includes the volleyball's three-dimensional spatial position and velocity vector.

[0010] In one specific embodiment, the player state vector and volleyball state vector output by the perception module are a complete quantitative description of the physical state of dynamic entities within the playing field at any given moment.

[0011] Specifically, the player state vector includes the player's three-dimensional spatial position, velocity vector, and body posture category. The three-dimensional spatial position refers to the real-time three-dimensional coordinates of the player's center of mass in a pre-calibrated world coordinate system with the center of the volleyball court as the origin. The velocity vector is a vector that includes both magnitude and direction, used to describe the player's instantaneous movement speed and direction in the world coordinate system. The velocity vector is obtained by differential calculation of the three-dimensional spatial position acquired in consecutive time frames. The body posture category is obtained by analyzing the player's real-time image through a posture estimation algorithm and classifying the current main body movements into a predefined category set. The category set includes: standing, moving, ready, jumping, spiking, blocking, or falling.

[0012] Specifically, the volleyball state vector includes the volleyball's three-dimensional spatial position and velocity vector. The definition of the three-dimensional spatial position is the same as that of the player's three-dimensional spatial position, and the definition of the velocity vector is the same as that of the player's velocity vector. These vectors are used to describe the instantaneous flight speed and direction of the volleyball in the world coordinate system.

[0013] The player state vectors and volleyball state vectors mentioned above together constitute the basic input data for all subsequent analysis modules.

[0014] In one specific embodiment, the player state vector acquired by the perception module includes the player's spatial position and velocity vector in the three-dimensional world coordinate system, as well as the body posture category obtained by the posture estimation algorithm; the volleyball state vector includes the volleyball's three-dimensional spatial position and velocity vector in the same coordinate system.

[0015] Specifically, the perception module also performs time series analysis on the player's state vector to extract the changing trends of motion features (such as average speed, acceleration extremes, and posture switching frequency), and infers the player's physiological indicators or fatigue level based on a preset exercise physiology model.

[0016] In one specific embodiment, the adaptive potential field kernel module includes an offline training phase and an online application phase. In the offline training phase, the adaptive potential field kernel module uses machine learning algorithms to learn from a large amount of historical match data of a specified player to construct a parameterized kernel model that can represent the player's ability and habits. The parameterized kernel model establishes a mapping relationship between the player's real-time state (including body posture, physiological indicators, or fatigue) and the potential field model parameters. In the online application phase, the adaptive potential field kernel module receives the player's real-time state from the perception module and uses the trained kernel model to dynamically and in real-time generate the optimal model parameters for the potential field generation module to call.

[0017] In one specific embodiment, the potential field generation module is further configured to generate a support potential field for quantifying the ability to organize an attack and the selection space. Correspondingly, the potential field coupling module is further configured to incorporate the support potential field into the coupling calculation to generate the net opportunity potential field. The support potential field is mainly generated by the setter, and its amplitude and range comprehensively reflect the potential of the setter to organize an effective attack under the current situation.

[0018] Specifically, the potential field coupling module is further configured to incorporate the support potential field, as a positive contribution representing the potential of one's own offensive organization, into the coupling calculation to generate the net opportunity potential field.

[0019] The support potential field is mainly generated by the setter. The amplitude and spatial range comprehensively reflect the potential of the setter to organize an effective attack and the range of tactical options in the current round. The generation of the support potential field is based on real-time quantitative analysis of at least one of the following key technical factors: the quality of the first pass, the movement state of the setter, and the relative position of the setter and the volleyball.

[0020] Specifically, the quantification of the quality of the first pass is achieved through the following steps: First, the system defines an optimal setting area in three-dimensional space in front of the net based on the setter's habits and standard tactical position; second, the volleyball state vector output by the sensing module is used in a physical motion model to predict the flight trajectory and final landing point of the first pass; finally, the system calculates the spatial deviation between the predicted landing point and the center point of the optimal setting area. The smaller the deviation, the higher the quality of the first pass, and the higher the amplitude of the generated support potential field.

[0021] Simultaneously, the system analyzes the setter's movement status, including the magnitude and direction of the velocity vector. A setter in a stationary or slow, stable movement state is considered to have a higher ability to organize attacks, thus enhancing the amplitude of the support field. Conversely, a setter in a high-speed, unstable state will weaken the amplitude of the support field.

[0022] In addition, the relative spatial position of the setter and the volleyball is also taken into account in the calculation. A situation in which the volleyball is in a favorable position in front of the setter's body will generate a wider and better-shaped support field. The spatial distribution of the support field represents the setter's available space to pass the ball to various attack points. On the other hand, a situation that forces the setter to make back or side adjustments will result in a narrower and more irregular support field, which objectively reflects that the tactical choices are restricted.

[0023] Ultimately, the potential field determined by these factors directly quantifies the potential of this offensive play, while the range and shape of the spatial distribution visually represent the areas where the setter can select to deliver high-quality passes, providing an objective data basis for evaluating subsequent offensive tactical choices.

[0024] In one specific embodiment, the generation of the support potential field is based on at least one of the following: the quality of the first pass, the speed of the setter's movement, and the relative position of the setter and the volleyball. The relative position of the setter and the volleyball is calculated in real time based on the player state vector and the volleyball state vector.

[0025] In one specific embodiment, the process of quantifying the quality of the first pass is as follows: First, an optimal setting area is defined based on the current position of the setter; then, the expected landing point of the first pass is calculated by physically modeling and predicting the motion trajectory represented by the volleyball state vector; finally, the spatial deviation between the expected landing point and the optimal setting area is calculated, and the deviation value is inversely proportional to the quantification result of the first pass quality.

[0026] In one specific embodiment, the potential field generation module generates the attack potential field through an anisotropic Gaussian function model.

[0027] Offensive momentum It can be represented by the following formula: ; in: The offensive potential field is represented by a scalar function, and its output value is expressed in two-dimensional coordinates on the court. place, time At that time, the potential attack threat intensity posed by the attacker is indicated by a higher value, which means a greater potential attack threat. This represents a two-dimensional coordinate vector of the stadium. It is a two-dimensional vector, for example... This indicates a specific location on the surface of the court. This represents a time variable, indicating the current moment. This represents the summation operator, which is used to sum a set. All eligible players The resulting individual offensive momentum is accumulated; Represents the set of attacking players, containing unique identifiers for all players currently in an offensive state; Indicates player At any moment The offensive threat amplitude is a scalar value that quantifies the player's offensive threat level. The attack strength that can be generated at the current moment The value depends on the player's real-time posture, relative position to the volleyball, and the player's individual ability; Represents the natural exponential function; Indicates player At any moment The two-dimensional position vector, this is a two-dimensional vector, representing the player's position. The two-dimensional projection coordinates on the field represent the coordinates of the players. The spatial center point of the generated individual offensive potential field; The transpose operator for a vector or matrix; Indicates player At any moment of The covariance matrix, a real symmetric positive definite matrix, is part of the parameters of the Gaussian model of the offensive potential field. Its elements determine the players'... The spatial distribution shape, expansion direction, and influence range of the generated individual offensive potential field are represented by the principal axis direction of the matrix, which characterizes the player. The main directionality of offensive threats; The inverse operator for matrices.

[0028] In one specific embodiment, the process by which the spatiotemporal analysis module generates the tactical momentum field includes calculating the partial derivative of the net opportunity potential field with respect to time, and determining a time component to characterize the rate of change of tactical opportunities over time.

[0029] The specific calculation method is as follows: ; in: The time component of the tactical momentum field is a scalar function whose output value is represented in two-dimensional coordinates on the field. place, time At that time, the rate of change of the net opportunity potential field over time, a positive value indicates that tactical opportunities are increasing, and a negative value indicates that tactical opportunities are decreasing; The net opportunity potential field is a scalar function whose output value is represented in two-dimensional coordinates on the field. place, time At that time, the net tactical opportunity distribution value after the interaction between offensive opportunities and defensive suppression on the field; For time Operators for calculating partial derivatives.

[0030] And calculate the gradient of the net opportunity potential field in spatial coordinates, and determine a spatial vector component to characterize the direction in which tactical opportunities grow the fastest in space: ; in: For the spatial vector components of the tactical momentum field, this is a two-dimensional vector function whose output vector is represented in two-dimensional coordinates on the field. place, time At that time, the direction in which the net opportunity potential field grows fastest in space, and the maximum rate of growth (i.e., the magnitude of the gradient). The gradient operator is defined as the vector formed by the partial derivatives of the function with respect to the coordinates in two-dimensional space for a scalar function. Net opportunity potential field to space The partial derivatives of the coordinates represent the partial derivatives of the coordinates. The rate of change of the net opportunity potential field along the axial direction; Net opportunity potential field to space The partial derivatives of the coordinates represent the partial derivatives of the coordinates. The rate of change of the net opportunity potential field along the axial direction.

[0031] In one specific embodiment, the decision support module is configured to visualize the scalar field of net opportunity potential field as a real-time heatmap; and to overlay the vector field of tactical momentum field on the real-time heatmap in the form of vector arrows, wherein the direction of the arrows indicates the direction of the spatial vector component, and the length or color of the arrows indicates the magnitude of the time component.

[0032] In one specific embodiment, the decision support module is configured to fuse and visualize the net opportunity potential field and the tactical momentum field.

[0033] The process of integrating visualization output first involves processing the net opportunity potential field. The decision support module discretizes the two-dimensional plane of the volleyball court into a grid with a preset resolution. For each cell in the grid, the system obtains the corresponding net opportunity potential field scalar value and converts it into a specific color value according to a preset color mapping table. The color mapping table maps high net opportunity potential fields to warm colors and low net opportunity potential fields to cool colors. Finally, a real-time updated heat map is generated on the display terminal to intuitively present the spatial distribution of tactical opportunities on the court.

[0034] The decision support module further displays the tactical momentum field as a vector arrow on the real-time heat map. In one embodiment, the vector arrow is drawn at the center point of the grid or sampled at preset intervals. The direction of each vector arrow is set to be consistent with the direction of the spatial vector component of the tactical momentum field at the center point of the grid, thereby indicating the direction in which the net opportunity potential field at this location grows fastest in space.

[0035] The visual attributes of the vector arrow are used to characterize the magnitude of the time component of the tactical momentum field. In one embodiment, the length of the arrow is proportional to the absolute value of the time component; a longer arrow indicates a faster rate of change in tactical opportunities at that location. In another embodiment, the color of the arrow is used to characterize the magnitude and sign of the time component. For example, a preset color spectrum maps positive time components, representing increasing opportunities, to the red family; and negative time components, representing decreasing opportunities or increasing threats, to the blue family, with the color saturation proportional to the absolute value of the time component.

[0036] This invention provides a real-time data analysis system for volleyball matches. It has the following beneficial effects: 1. This invention, by setting up a potential field generation module and a potential field coupling module, can generate an offensive potential field that quantifies offensive threats and a defensive potential field that quantifies defensive coverage capabilities based on the real-time state vectors of players and volleyballs. It also couples the potential fields into a net opportunity potential field that characterizes the distribution of tactical opportunities. This approach transforms the game situation, which relies on experience-based judgment, into an objective, continuous, and quantifiable data field. This enables real-time analysis of the fleeting tactical games in the game, solving the problem that existing technologies can only perform post-event statistics and cannot quantitatively evaluate the real-time tactical situation.

[0037] 2. By setting an adaptive potential field kernel module, this invention can dynamically adjust the model parameters used to generate the potential field based on the historical match data, physiological indicators, fatigue levels, and other real-time status of specified players. This setting enables the potential field model to reflect the technical characteristics, habits, and state fluctuations caused by changes in physical strength during the match, thereby improving the accuracy and realism of the tactical situation model generated by the entire analysis system and avoiding the problem of analysis results not matching the actual situation caused by using a fixed parameter model.

[0038] 3. By setting up a spatiotemporal analysis module, this invention can calculate the tactical momentum field based on the time series of the net opportunity potential field. The tactical momentum field, through the temporal gradient and spatial gradient, respectively characterizes the rate of change and evolution direction of tactical opportunities, thereby improving the system's analytical capabilities from describing static situations to predicting dynamic trends. This enables the system to not only identify the current opportunity distribution but also reveal the generation and development of opportunity windows, providing data support for tactical decision-making. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the system functional module structure of the present invention; Figure 2 This is a schematic diagram of the system hardware deployment environment of the present invention; Figure 3 This is a schematic diagram of the potential field generation and coupling process of the present invention; Figure 4 This is a schematic diagram of the tactical momentum field process of the present invention; Figure 5 This is a flowchart of the data processing of the perception and kernel module of the present invention. Detailed Implementation

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Please see the appendix Figure 1 - Appendix Figure 3 This invention provides a real-time data analysis system for volleyball matches, including a perception module, an adaptive potential field kernel module, a potential field generation module, a potential field coupling module, a spatiotemporal analysis module, and a decision support module.

[0042] In this embodiment, the overall data flow process of the system involves the sensing module collecting data from the competition venue to obtain raw physical state data. This data is simultaneously transmitted to the adaptive potential field kernel module and the potential field generation module. The adaptive potential field kernel module dynamically generates model parameters based on the received real-time state and transmits them to the potential field generation module.

[0043] In this embodiment, the overall data flow process of the system starts from the perception module. The perception module collects real-time data of the competition field and obtains and outputs two core raw physical state data: one is the player state vector of at least one player, and the other is the volleyball state vector. The player state vector specifically includes the player's three-dimensional spatial position, velocity vector and body posture category.

[0044] Subsequently, the data stream branches, and the player state vector and volleyball state vector are directly transmitted to the potential field generation module as the basic physical states on which to construct each potential field. At the same time, the player state vector is also synchronously transmitted to the adaptive potential field kernel module.

[0045] After receiving the player's real-time status, the adaptive potential field kernel module dynamically generates a set of model parameters using its pre-established internal mapping relationship. These model parameters are used to personalize and dynamically adjust the shape of the offensive or defensive potential field that the player will generate. For example, they determine the element values ​​of the covariance matrix used to characterize the directionality and range of influence of the offensive threat. Finally, this set of dynamically generated model parameters is transmitted from the adaptive potential field kernel module to the potential field generation module.

[0046] The potential field generation module combines physical state data from the perception module and model parameters from the adaptive potential field kernel module to generate multiple independent tactical potential fields, which are then transmitted to the potential field coupling module.

[0047] The input of the potential field generation module is connected to the output of the sensing module and the adaptive potential field kernel module, respectively. Its function is to receive physical state data and model parameters, and generate multiple independent tactical potential fields to represent different tactical functions. These generated tactical potential fields are then transmitted as a whole to the potential field coupling module for further processing.

[0048] Specifically, the generation process includes: An offensive potential field is generated to quantify the offensive threat. This offensive potential field is implemented through an anisotropic Gaussian function model. Several parameters of this model, including an amplitude representing the intensity of the threat, a spatial center corresponding to the current position of the offensive player, and a covariance matrix that determines the shape of the threat distribution in space, the range of influence, and the main direction of action, are jointly determined by the player state vector output by the perception module and the model parameters output by the adaptive potential field kernel module.

[0049] Simultaneously, a defensive potential field is generated to quantify the defender's spatial coverage capability. This defensive potential field can be realized through an isotropic model, whose magnitude represents the defensive coverage capability and whose radius represents the range of influence. These parameters are also determined by the real-time status of the defensive player and the model parameters generated by the adaptive potential field kernel module.

[0050] In addition, a support potential field is generated to quantify the organization's ability to launch a second attack. The generation of this support potential field is based on a comprehensive calculation of at least one of the following: the quantification of the quality of the first pass, the setter's movement speed, and the relative positional relationship between the setter and the volleyball.

[0051] The potential field coupling module performs coupling calculations on multiple received tactical potential fields to generate a unified net opportunity potential field. This net opportunity potential field is continuously transmitted as time series data to the spatiotemporal analysis module and the decision support module.

[0052] The spatiotemporal analysis module processes the time series of the net opportunity potential field, calculates the tactical momentum field, and transmits the results to the decision support module.

[0053] Finally, the decision support module integrates data from the potential field coupling module and the spatiotemporal analysis module to generate a visual output.

[0054] Specifically, the perception module is designed to acquire in real time the player state vector of at least one player and the volleyball state vector of the volleyball during the match. The player state vector includes the player's three-dimensional spatial position, velocity vector, and body posture category; the volleyball state vector includes the volleyball's three-dimensional spatial position and velocity vector.

[0055] The adaptive potential field kernel module is connected to the output of the perception module. Its function is to dynamically generate model parameters for constructing offensive and defensive potential fields based on the real-time status of a specified player obtained from the perception module.

[0056] The potential field generation module has its input terminals connected to the output terminals of the perception module and the adaptive potential field kernel module, respectively. Its function is to generate an offensive potential field that represents offensive threat, a defensive potential field that represents defensive coverage capability, and a support potential field that represents the ability to organize offense, based on the player's state vector, the volleyball's state vector, and the model parameters.

[0057] The potential field coupling module is connected to the output of the potential field generation module. Its function is to generate a net opportunity potential field that represents the distribution of tactical opportunities on the field by performing coupled calculations on the offensive potential field, defensive potential field, and support potential field.

[0058] This coupling calculation can be characterized by the following formula: ; in: In the two-dimensional coordinates of the court and time Net opportunity potential value; It is the offensive momentum field value; It is the defensive field value; It is the field value of the support potential; These are preset weighting coefficients or dynamically adjusted weighting coefficients based on the racing stage.

[0059] Specifically, in one particular embodiment, the dynamic adjustment of the weighting coefficients is achieved through a finite state machine. This state machine predefines multiple game stage states (e.g., service set, first attack, counter-attack). The system identifies the current game stage state by analyzing the volleyball state vector and the player state vector, and automatically calls the preset weighting coefficient combination corresponding to that state to weight the calculation of the net chance potential field.

[0060] The spatiotemporal analysis module is connected to the output of the potential field coupling module. Its function is to calculate a tactical momentum field based on the change of the net opportunity potential field in a continuous time series. This tactical momentum field consists of a time component and a spatial vector component.

[0061] The decision support module, whose input is connected to the output of the potential field coupling module and the spatiotemporal analysis module respectively, is used to visualize one or more of the net opportunity potential field and tactical momentum field.

[0062] See attached document Figure 5 The perception module acquires multi-view video streams through a multi-camera array deployed around the competition venue. The computer vision algorithm integrated within the module processes the video streams. First, it identifies and continuously tracks all players and volleyballs through target detection and tracking algorithms. Then, it uses a 3D reconstruction algorithm to map the 2D image coordinates to a pre-calibrated 3D world coordinate system, thereby outputting the 3D spatial position of each player and the 3D spatial position of the volleyball.

[0063] By calculating the difference in position changes within consecutive time frames, the velocity vectors of the player and the volleyball are obtained. Simultaneously, a pose estimation algorithm processes the player image and outputs its body pose category. The perception module also analyzes the time series of the player's state vectors to extract motion features and, based on a sports physiology model, infers the player's physiological indicators or fatigue level. Specifically, the inference process includes: within a sliding time window, calculating the temporal statistical features (e.g., mean, variance, kurtosis) of velocity and acceleration in the player's state vector, as well as the frequency domain features (e.g., frequency of posture switching, duration distribution of different postures) of the body posture category sequence; using these extracted motion features as input, feeding them into a pre-trained regression model (e.g., gradient boosting decision tree or recurrent neural network), the output of which is a quantitative estimate of the player's current physiological indicators or fatigue state.

[0064] The adaptive potential field kernel module includes an offline training process and an online application process. During the offline training process, the module uses machine learning algorithms to learn from the historical match data of a specified player in order to build a parameterized kernel model. This model establishes a mapping relationship between the player's real-time state and its potential field model parameters.

[0065] Specifically, the machine learning algorithm can be Gaussian process regression or feedforward neural network. The input of the algorithm is the player's real-time state vector, and the output is the model parameters used to generate the player's offensive or defensive potential field, such as the element values ​​of the covariance matrix in the offensive potential field model.

[0066] During online application, this module receives the real-time status of players from the perception module and uses the trained kernel model to generate model parameters in real time.

[0067] The potential field generation module receives the state vector from the perception module and the model parameters from the adaptive potential field kernel module to generate multiple independent potential fields.

[0068] The generation of the offensive potential field is achieved through an anisotropic Gaussian function model.

[0069] Offensive momentum It can be represented by the following formula: ; in: Representing the offensive potential field, this is a scalar function whose output value is expressed in two-dimensional coordinates on the court. place, time At that time, the strength of the potential attack threat posed by the attacker; the higher the value, the greater the potential attack threat. This represents a two-dimensional coordinate vector of the stadium. It is a two-dimensional vector, for example... This indicates a specific location on the surface of the court. This represents a time variable, indicating the current moment. This represents the summation operator, which is used to sum a set. All eligible players The resulting individual offensive momentum is accumulated; Represents the set of attacking players, containing unique identifiers for all players currently in an offensive state; Indicates player At any moment The offensive threat amplitude is a scalar value that quantifies the player's offensive threat level. The attack strength that can be generated at a given moment depends on the player's real-time posture, relative position to the volleyball, and the player's individual ability. Represents the natural exponential function; Indicates player At any moment The two-dimensional position vector, this is a two-dimensional vector, representing the player's position. The two-dimensional projection coordinates on the field represent the coordinates of the players. The spatial center point of the generated individual offensive potential field; The transpose operator for a vector or matrix; Indicates player At any moment of The covariance matrix, a real symmetric positive definite matrix, is part of the parameters of the Gaussian model of the offensive potential field. Its elements determine the players'... The spatial distribution shape, expansion direction, and influence range of the generated individual offensive potential field are represented by the principal axis direction of the matrix, which characterizes the player's... The main directionality of offensive threats; The inverse operator for matrices.

[0070] The defensive potential field is generated through an isotropic Gaussian function model, the magnitude of which and the radius of influence are determined by the player's real-time state and the model parameters generated by the adaptive potential field kernel module.

[0071] The generation of the support potential field is based on at least one of the following: the quality of the first pass, the speed of the setter's movement, and the relative position of the setter and the volleyball. The process of quantifying the quality of the first pass involves defining an optimal setter zone based on the setter's current position; then, by physically modeling and predicting the trajectory represented by the volleyball's state vector, the expected landing point of the first pass is calculated; finally, the spatial deviation between the expected landing point and the optimal setter zone is calculated, and this deviation value is inversely proportional to the quantification result of the first pass quality.

[0072] The spatiotemporal analysis module processes the time series of the net opportunity potential field to generate a tactical momentum field.

[0073] The process includes: calculating the partial derivative of the net opportunity potential field with respect to time to determine a time component that characterizes the rate of change of tactical opportunities over time.

[0074] ; in: The time component of the tactical momentum field is a scalar function whose output value is represented in two-dimensional coordinates on the field. place, time At that time, the rate of change of the net opportunity potential field over time, a positive value indicates that tactical opportunities are increasing, and a negative value indicates that tactical opportunities are decreasing; The net opportunity potential field is a scalar function whose output value is represented in two-dimensional coordinates on the field. place, time At that time, the net tactical opportunity distribution value after the interaction between offensive opportunities and defensive suppression on the field; For time Operators for calculating partial derivatives.

[0075] And calculate the gradient of the net opportunity potential field in spatial coordinates, and determine a spatial vector component to characterize the direction in which tactical opportunities grow the fastest in space: ; in: For the spatial vector components of the tactical momentum field, this is a two-dimensional vector function whose output vector is represented in two-dimensional coordinates on the field. place, time At that time, the direction in which the net opportunity potential field grows fastest in space, and the maximum rate of that growth (i.e., the magnitude of the gradient). The gradient operator is defined as the vector formed by the partial derivatives of the function with respect to the coordinates in two-dimensional space for a scalar function. Net opportunity potential field to space The partial derivatives of the coordinates represent the partial derivatives of the coordinates. The rate of change of the net opportunity potential field along the axial direction; Net opportunity potential field to space The partial derivatives of the coordinates represent the partial derivatives of the coordinates. The rate of change of the net opportunity potential field along the axial direction.

[0076] The decision support module is configured to visualize the scalar field of net opportunity potential field as a real-time heatmap; and to overlay the vector field of tactical momentum field on the real-time heatmap as vector arrows, where the direction of the arrows indicates the direction of the spatial vector components, and the length or color of the arrows indicates the magnitude of the time components.

[0077] See attached document Figure 1 - Appendix Figure 5The system of the present invention can be deployed in a hardware environment that includes a data acquisition subsystem, a data processing subsystem, and a data presentation subsystem.

[0078] The data acquisition subsystem includes multiple high-speed cameras configured to simultaneously acquire multi-view video streams of the competition venue and connects to the data processing subsystem via a data communication link. This subsystem provides raw data input to the sensing module.

[0079] The data processing subsystem may be one or more servers, each including at least one processor, memory, graphics processing unit (GPU), and network interface. The memory is used to store computer program instructions, which, when executed by the processor, implement some functions of the perception module, as well as all functions of the adaptive potential field kernel module, potential field generation module, potential field coupling module, spatiotemporal analysis module, and decision support module. The graphics processing unit (GPU) is used to accelerate large-scale parallel computing tasks in computer vision algorithms and potential field calculations.

[0080] The data presentation subsystem can be a display terminal connected to the data processing subsystem, used to receive and display the visualization output generated by the decision support module, such as real-time heatmaps and overlaid vector arrows.

[0081] In a specific deployment scheme, some functions of the perception module, especially the preliminary processing of video streams and extraction of state vectors, can be executed on edge computing nodes deployed close to the camera array to reduce data transmission bandwidth. Meanwhile, computationally intensive modules such as the adaptive potential field kernel module, potential field generation module, potential field coupling module, and spatiotemporal analysis module are centrally deployed in the central data processing subsystem to utilize its centralized computing resources.

Claims

1. A volleyball match real-time data analysis system, characterized in that, include: The perception module is configured to acquire the player state vector of at least one player and the volleyball state vector of the volleyball in real time during the match. The adaptive potential field kernel module is configured to dynamically generate model parameters for constructing offensive and defensive potential fields based on the real-time status of specified players. The potential field generation module is configured to generate an offensive potential field and a defensive potential field based on the player state vector, the volleyball state vector and the model parameters. The potential field coupling module is configured to generate a net opportunity potential field that characterizes the distribution of tactical opportunities on the field by performing coupled calculations on the offensive potential field and the defensive potential field. The spatiotemporal analysis module is configured to calculate a tactical momentum field based on the changes of the net opportunity potential field over a continuous time series. The decision support module is configured to visualize one or more of the net opportunity potential field and the tactical momentum field.

2. The volleyball match real-time data analysis system according to claim 1, characterized in that, The player state vector includes the player's three-dimensional spatial position, velocity vector, and body posture category, while the volleyball state vector includes the volleyball's three-dimensional spatial position and velocity vector.

3. The volleyball match real-time data analysis system according to claim 1, characterized in that, The sensing module is also configured to: Based on the characteristics of the player's state vector changing over time, at least one player's physiological indicators or fatigue level can be inferred.

4. The volleyball match real-time data analysis system according to claim 1, characterized in that, The adaptive potential field kernel module is also configured to: Offline training is performed based on the historical match data of the specified player to learn the mapping relationship between the real-time state and the model parameters. The real-time state includes the player's physiological indicators or fatigue level.

5. The volleyball match real-time data analysis system according to claim 1, characterized in that, The potential field generation module is further configured to generate a support potential field for quantifying the organization's offensive capabilities and choice space; the potential field coupling module is further configured to incorporate the support potential field into the coupling calculation to generate the net opportunity potential field.

6. The volleyball match real-time data analysis system according to claim 5, characterized in that, The generation of the support potential field is based on at least one of the following: the quality of the first pass, the speed of the setter's movement, and the relative position of the setter and the volleyball. The relative position of the setter and the volleyball is calculated based on the player's state vector and the volleyball's state vector.

7. The volleyball match real-time data analysis system according to claim 6, characterized in that, The quality of the first pass is determined quantitatively by calculating the spatial deviation between the landing point of the volleyball trajectory represented by the volleyball state vector and an optimal setting area preset based on the setter's position.

8. The volleyball match real-time data analysis system according to claim 1, characterized in that, The potential field generation module generates the offensive potential field through an anisotropic Gaussian function model, wherein a covariance matrix of the Gaussian function model is used as part of the model parameters to characterize the directionality of the offensive threat.

9. The volleyball match real-time data analysis system according to claim 1, characterized in that, The generation of the tactical momentum field includes: calculating the time gradient of the net opportunity potential field and determining a time component to characterize the rate of change of tactical opportunities; and calculating the spatial gradient of the net opportunity potential field and determining a spatial vector component to characterize the direction of growth of tactical opportunities.

10. The volleyball match real-time data analysis system according to claim 1, characterized in that, The decision support module is specifically configured to output the net opportunity potential field in the form of a real-time heatmap, and to overlay the tactical momentum field on the real-time heatmap in the form of vector arrows.

Images