A sports badminton skill and tactic style recognition and confrontation simulation system based on big data analysis

Abstract

The application provides a sports badminton skill and tactics style recognition and confrontation simulation system based on big data analysis, which is composed of the following four layer groups: the first layer group: data acquisition and processing layer; the second layer group: analysis and recognition layer; the third layer group: confrontation simulation and interaction layer; the fourth layer group: application output layer; and further includes the following three core modules: module one: skill and tactics style recognition module; module two: confrontation simulation module; module three: interaction and decision support module, the application will construct a complete system through the four layer groups of data acquisition and processing layer, analysis and recognition layer, confrontation simulation and interaction layer, application output layer and the three core modules of skill and tactics style recognition module, confrontation simulation module and interaction and decision support module, so that it can effectively combine big data and the analyzed and simulated confrontation of the entered video, so that the athletes can effectively respond to different skill and tactics styles according to the simulated confrontation.

Description

Technical Field

[0001] This invention relates to the field of sports tactical recognition and simulation technology, and more specifically to a sports badminton tactical style recognition and confrontation simulation system based on big data analysis. Background Technology

[0002] Badminton tactical style refers to a relatively stable system of playing methods and combat strategies formed by athletes or teams in a competition, based on their own characteristics and understanding of the laws of badminton. Therefore, athletes can record relevant videos during training and watch them repeatedly to study specific techniques and think about countermeasures, so as to formulate subsequent training plans and improve their badminton skills. However, simply recording videos and repeatedly watching them to create customized training plans can weaken certain details that are generated by human thought, leading to defeat in subsequent actual combat due to neglect of some key details. To this end, we will combine the gradual development of modern technology and use big data and AI analysis for identification and simulation. As a result, we will propose a sports badminton tactical style identification and confrontation simulation system based on big data analysis to improve the overall training intensity and accuracy of training methods for athletes. Summary of the Invention

[0003] 9. The technical solution adopted by this invention to achieve its technical objective is as follows: A sports badminton tactical style recognition and confrontation simulation system based on big data analysis will consist of the following four layers: First layer: Data acquisition and processing layer; Second layer: Analysis and identification layer; The third layer: adversarial simulation and interaction layer; Fourth layer group: Application output layer; It also includes the following three core modules: Module 1: Tactical Style Recognition Module; Module Two: Adversarial Simulation Module; Module 3: Interaction and Decision Support Module; The system combines four layers and three core modules to form a complete sports badminton tactical style recognition and confrontation simulation system.

[0004] As a further improvement of the present invention, the data acquisition and processing layer of the first layer group includes: match video input, computer vision processing (YOLO object detection / MediaPipe pose estimation), and feature extraction (player trajectory, hitting action, tactical events).

[0005] As a further improvement of the present invention, the analysis and identification layer of the second layer group includes: technical and tactical style identification, machine learning classification model, tactical knowledge graph construction, spatiotemporal event structuring, and key indicator calculation (out-of-position state, movement distance, and smash speed).

[0006] As a further improvement of the present invention, the adversarial simulation and interaction layer of the third layer group includes: a strong learning simulation environment (ShuttleEnv, intelligent agent battle), scenario-based training generation (CGAN dynamic scenario), and natural language interactive query (RAG+LLM question answering system).

[0007] As a further improvement of the present invention, the application output layer of the fourth layer group includes: coach and athlete decision support, personalized tactical training program generation, and intelligent viewing visualization.

[0008] As a further improvement of the present invention, the tactical style recognition module of module one performs: visual perception, action recognition, and style classification. Visual perception: Use target detection algorithms such as YOLO to locate players and badminton shuttlecocks, use MediaPipe to estimate human pose, and extract the skeletal key points and movement trajectory of players; Action recognition: Using deep learning models such as 3D-CNN, LSTM or Video Swin Transformer, specific hitting action types (such as smash, drop shot, high clear) are identified. Studies show that the optimized system can achieve an accuracy of over 94% in action recognition. Style classification: Based on the extracted tactical event sequences (such as the landing point of the ball and the movement path), the system uses machine learning models (such as PFA-INT-SVM) to classify athletes into specific tactical styles (such as drop shot attack, offensive, and counter-attacking). The system can achieve a tactical prediction accuracy of 85.4%, which is significantly higher than the 75.1% of traditional methods. The system's "eyes" and "analysis brain" are responsible for "understanding" the competition from the videos.

[0009] As a further improvement of the present invention, the adversarial simulation module of module two performs the following tasks: tactical knowledge graph, reinforcement learning environment, and dynamic scene generation. Tactical Knowledge Graph: Construct a multi-dimensional tactical knowledge base that includes spatial trajectory, hitting pattern, and confrontation strategy, and structure tactical knowledge for subsequent reasoning. Reinforced learning environment: Platforms like ShuttleEnv build probabilistic models based on real match data from top athletes to simulate round dynamics and train AI agents with different playing styles. Players can compete against these AIs in a simulated environment to explore the best strategies for dealing with opponents of different styles. Dynamic scene generation: By using generative AI (such as conditional generative adversarial networks CGAN), diverse adversarial scenes can be automatically generated according to training requirements. The difficulty and policy complexity can be dynamically adjusted through the proximal policy optimization (PPO) algorithm to achieve true "personalized training". The system's "sand table simulation room" allows you to battle against AI opponents of different styles in a virtual environment.

[0010] As a further improvement of the present invention, the interaction and decision support module of module three performs the following tasks: natural language question answering, key parameter visualization, and training prescription generation. Natural Language Question Answering: With the help of Retrieval Enhanced Generation (RAG) technology, coaches can ask questions in everyday language, such as "Why do I always lose points on the backhand side in the backcourt?" The system can retrieve the tactical knowledge base and provide a text analysis report with data support. Visualization of key parameters: The system can overlay key indicators such as the athlete's movement distance, maximum smash speed, jump height, and offensive and defensive paths in real time during live broadcasts or replays, making tactical analysis clear at a glance; Training prescription generation: Based on style analysis of athletes and opponents, the system can automatically generate targeted training plans or competition strategy suggestions; The system's "dialogue interface" makes complex analysis results readily accessible.

[0011] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention constructs a complete system through four layers: data acquisition and processing layer, analysis and identification layer, adversarial simulation and interaction layer, and application output layer, as well as three core modules: technical and tactical style identification module, adversarial simulation module, and interaction and decision support module. This system can effectively combine big data and recorded video for analysis and simulated adversarial training. As a result, athletes can effectively respond to different technical and tactical styles based on simulated adversarial training. At the same time, the generated simulation environment will be used for tactical deduction and training, thereby improving the training effect of athletes.

[0012] 2. This invention replaces the traditional reliance on coaches' subjective experience for tactical analysis with data-driven, quantifiable, and interactive scientific decision-making, thereby enabling precise analysis of all aspects of tactics and ensuring that no key details are overlooked. Attached Figure Description

[0013] Figure 1 This diagram illustrates the four layers of a sports badminton tactical style recognition and confrontation simulation system based on big data analysis.

[0014] Figure 2This diagram illustrates the three core modules of a sports badminton tactical style recognition and confrontation simulation system based on big data analysis.

[0015] Figure 3 This is a schematic diagram of a technical and tactical style recognition module.

[0016] Figure 4 This is a schematic diagram of an adversarial simulation module.

[0017] Figure 5 This is a schematic diagram of an interaction and decision support module. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings: Example: Figures 1 to 5 As shown: This invention provides a sports badminton tactical style recognition and confrontation simulation system based on big data analysis, which consists of the following four layers: First layer: Data acquisition and processing layer; Second layer: Analysis and identification layer; The third layer: adversarial simulation and interaction layer; Fourth layer group: Application output layer; It also includes the following three core modules: Module 1: Tactical Style Recognition Module; Module Two: Adversarial Simulation Module; Module 3: Interaction and Decision Support Module; The system combines four layers and three core modules to form a complete sports badminton tactical style recognition and confrontation simulation system.

[0019] The data acquisition and processing layer of the first layer includes: match video input, computer vision processing (YOLO object detection / MediaPipe pose estimation), and feature extraction (player trajectory, hitting action, tactical events).

[0020] The analysis and identification layer of the second group includes: tactical style identification, machine learning classification model, tactical knowledge graph construction, spatiotemporal event structuring, and key indicator calculation (out-of-position state, movement distance, and smash speed).

[0021] The adversarial simulation and interaction layer of the third layer includes: a strong learning simulation environment (ShuttleEnv, agent battle), scenario-based training generation (CGAN dynamic scenario), and natural language interactive query (RAG+LLM question answering system).

[0022] The application output layer of the fourth layer group includes: coach and athlete decision support, personalized tactical training program generation, and smart viewing visualization.

[0023] Among them, the tactical style recognition module of module one performs the following tasks: visual perception, action recognition, and style classification. Visual perception: Use target detection algorithms such as YOLO to locate players and badminton shuttlecocks, use MediaPipe to estimate human pose, and extract the skeletal key points and movement trajectory of players; Action recognition: Using deep learning models such as 3D-CNN, LSTM or Video Swin Transformer, specific hitting action types (such as smash, drop shot, high clear) are identified. Studies show that the optimized system can achieve an accuracy of over 94% in action recognition. Style classification: Based on the extracted tactical event sequences (such as the landing point of the ball and the movement path), the system uses machine learning models (such as PFA-INT-SVM) to classify athletes into specific tactical styles (such as drop shot attack, offensive, and counter-attacking). The system can achieve a tactical prediction accuracy of 85.4%, which is significantly higher than the 75.1% of traditional methods. The system's "eyes" and "analysis brain" are responsible for "understanding" the competition from the videos.

[0024] Among them, the adversarial simulation module of module two performs the following tasks: tactical knowledge graph, reinforcement learning environment, and dynamic scene generation. Tactical Knowledge Graph: Construct a multi-dimensional tactical knowledge base that includes spatial trajectory, hitting pattern, and confrontation strategy, and structure tactical knowledge for subsequent reasoning. Reinforced learning environment: Platforms like ShuttleEnv build probabilistic models based on real match data from top athletes to simulate round dynamics and train AI agents with different playing styles. Players can compete against these AIs in a simulated environment to explore the best strategies for dealing with opponents of different styles. Dynamic scene generation: By using generative AI (such as conditional generative adversarial networks CGAN), diverse adversarial scenes can be automatically generated according to training requirements. The difficulty and policy complexity can be dynamically adjusted through the proximal policy optimization (PPO) algorithm to achieve true "personalized training". The system's "sand table simulation room" allows you to battle against AI opponents of different styles in a virtual environment.

[0025] Among them, the interaction and decision support module of module three performs the following tasks: natural language question answering, key parameter visualization, and training prescription generation. Natural Language Question Answering: With the help of Retrieval Enhanced Generation (RAG) technology, coaches can ask questions in everyday language, such as "Why do I always lose points on the backhand side in the backcourt?" The system can retrieve the tactical knowledge base and provide a text analysis report with data support. Visualization of key parameters: The system can overlay key indicators such as the athlete's movement distance, maximum smash speed, jump height, and offensive and defensive paths in real time during live broadcasts or replays, making tactical analysis clear at a glance; Training prescription generation: Based on style analysis of athletes and opponents, the system can automatically generate targeted training plans or competition strategy suggestions; The system's "dialogue interface" makes complex analysis results readily accessible.

[0026] The specific functions and operation procedures of this embodiment are as follows: In this invention, the sports badminton tactical style recognition and confrontation simulation system based on big data analysis will consist of the following four layers: First layer: Data acquisition and processing layer; Second layer: Analysis and identification layer; The third layer: adversarial simulation and interaction layer; Fourth layer group: Application output layer; It also includes the following three core modules: Module 1: Tactical Style Recognition Module; Module Two: Adversarial Simulation Module; Module 3: Interaction and Decision Support Module; Combining four layers and three core modules, a complete sports badminton tactical style recognition and confrontation simulation system is formed, creating an intelligent platform integrating computer vision, machine learning, and large-scale voice models. It can automatically extract athletes' movement and tactical data from match videos, identify different tactical styles, and generate interactive simulation environments for tactical deduction and training. Simultaneously, it can replace traditional coaches' subjective experience in tactical analysis, transforming it into data-driven, quantifiable, and interactive scientific decision-making. This creates an intelligent decision-making platform that deeply integrates AI and sports science, using computer vision to "understand" matches, machine learning to "comprehend" tactics, and reinforcement learning to "simulate" confrontations. Ultimately, it "communicates" with coaches and athletes through natural language. It is a key tool for transforming top-level competition experience into replicable and interactive digital assets, propelling badminton training and research from "experience-driven" to a new era of "data-intelligent-driven."

[0027] Any technical solution that achieves the above-mentioned technical effects by utilizing the technical solutions described in this invention, or by designing similar technical solutions by those skilled in the art under the inspiration of the technical solutions described in this invention, falls within the protection scope of this invention.

Claims

1. A sports badminton tactical style recognition and confrontation simulation system based on big data analysis will consist of the following four layers: First layer: Data acquisition and processing layer; Second layer: Analysis and identification layer; The third layer: adversarial simulation and interaction layer; Fourth layer group: Application output layer; It also includes the following three core modules: Module 1: Tactical Style Recognition Module; Module Two: Adversarial Simulation Module; Module 3: Interaction and Decision Support Module; The system combines four layers and three core modules to form a complete sports badminton tactical style recognition and confrontation simulation system.

2. The sports badminton tactical style recognition and confrontation simulation system based on big data analysis according to claim 1, characterized in that: The data acquisition and processing layer of the first layer group includes: match video input, computer vision processing (YOLO object detection / MediaPipe pose estimation), and feature extraction (player trajectory, hitting action, tactical events).

3. The sports badminton tactical style recognition and confrontation simulation system based on big data analysis according to claim 1, characterized in that: The analysis and identification layer of the second group includes: technical and tactical style identification, machine learning classification model, tactical knowledge graph construction, spatiotemporal event structuring, and key indicator calculation (out-of-position state, movement distance, and smash speed).

4. The sports badminton tactical style recognition and confrontation simulation system based on big data analysis according to claim 1, characterized in that: The adversarial simulation and interaction layer of the third layer includes: a strong learning simulation environment (ShuttleEnv, agent battle), scenario-based training generation (CGAN dynamic scenario), and natural language interactive query (RAG+LLM question answering system).

5. The sports badminton tactical style recognition and confrontation simulation system based on big data analysis according to claim 1, characterized in that: The application output layers of the fourth layer group include: decision support for coaches and athletes, generation of personalized tactical training plans, and visualization of smart viewing.

6. The sports badminton tactical style recognition and confrontation simulation system based on big data analysis according to claim 1, characterized in that: The tactical style recognition module of module one performs the following tasks: visual perception, action recognition, and style classification. Visual perception: Use target detection algorithms such as YOLO to locate players and badminton shuttlecocks, use MediaPipe to estimate human pose, and extract the skeletal key points and movement trajectory of players; Action recognition: Using deep learning models such as 3D-CNN, LSTM or Video Swin Transformer, specific hitting action types (such as smash, drop shot, high clear) are identified. Studies show that the optimized system can achieve an accuracy of over 94% in action recognition. Style classification: Based on the extracted tactical event sequences (such as the landing point of the ball and the movement path), the system uses machine learning models (such as PFA-INT-SVM) to classify athletes into specific tactical styles (such as drop shot attack, offensive, and counter-attacking). The system can achieve a tactical prediction accuracy of 85.4%, which is significantly higher than the 75.1% of traditional methods. The system's "eyes" and "analysis brain" are responsible for "understanding" the competition from the videos.

7. The sports badminton tactical style recognition and confrontation simulation system based on big data analysis according to claim 1, characterized in that: The adversarial simulation module in Module 2 performs tasks such as tactical knowledge graph generation, reinforcement learning environment, and dynamic scenario generation. Tactical Knowledge Graph: Construct a multi-dimensional tactical knowledge base that includes spatial trajectory, hitting pattern, and confrontation strategy, and structure tactical knowledge for subsequent reasoning. Reinforced learning environment: Platforms like ShuttleEnv build probabilistic models based on real match data from top athletes to simulate round dynamics and train AI agents with different playing styles. Players can compete against these AIs in a simulated environment to explore the best strategies for dealing with opponents of different styles. Dynamic scene generation: By using generative AI (such as conditional generative adversarial networks CGAN), diverse adversarial scenes can be automatically generated according to training requirements. The difficulty and policy complexity can be dynamically adjusted through the proximal policy optimization (PPO) algorithm to achieve true "personalized training". The system's "sand table simulation room" allows you to battle against AI opponents of different styles in a virtual environment.

8. The sports badminton tactical style recognition and confrontation simulation system based on big data analysis according to claim 1, characterized in that: The interaction and decision support module of module three performs the following tasks: natural language question answering, key parameter visualization, and training prescription generation. Natural Language Question Answering: With the help of Retrieval Enhanced Generation (RAG) technology, coaches can ask questions in everyday language, such as "Why do I always lose points on my backhand side in the backcourt?" The system can retrieve the tactical knowledge base and provide a text analysis report with data support. Visualization of key parameters: The system can overlay key indicators such as the athlete's movement distance, maximum smash speed, jump height, and offensive and defensive paths in real time during live broadcasts or replays, making tactical analysis clear at a glance; Training prescription generation: Based on style analysis of athletes and opponents, the system can automatically generate targeted training plans or competition strategy suggestions; The system's "dialogue interface" makes complex analysis results readily accessible.

Images