I don't know if you have noticed that in the wonderful semifinals of Sun Yingsha and Mimato Ito at the Tokyo Olympics, there was an intelligent big data analysis and decision-making system that helped the Chinese team monitor and analyze the whole process before and after the competition to provide support for winning the championship.
What is this system?
This system comes from the laboratory of Zhejiang University, in the table tennis field of the Tokyo Olympics, through the artificial intelligence platform built by 8 special cameras to monitor and analyze the whole process, it loyally records every serve, swing, run and other data of Mimato Ito. Combined with the analysis of the data of the past 8,000 historical matches and the on-site average speed of 100Mbps, the sports data is transmitted to the system to analyze the player's skills and tactics, the identification of the state of peaks and troughs, the advantages and characteristics, etc., to find winning tactical strategies, and push them to coaches to assist athletes in decision-making, so that coaches and athletes can "know themselves and know each other, and fight a hundred battles".
With such a powerful boost, Sun Yingsha on the field quickly curbed Ito's technical play and finally won 4-0. In the post-match interview, Ito also had to admit that his level only played two or three percent.
Tokyo Olympics Sun Yingsha and Ito Mimato semifinals Photo/Osports
It can be seen that science and technology to help sports have become the lever and grasp for the sustainable development of mainland competitive sports, and their impact on the results of competitions is increasing. Establishing a big data and intelligent system is an effective means for science and technology to empower sports and carry out digital science training.
What is Digital Science Training?
Digital science training mainly focuses on the technical and tactical, physical and psychological training of outstanding athletes, and the laboratory provides training monitoring, technical and tactical video analysis and competition data reporting for the training and competition of outstanding athletes through artificial intelligence technologies such as data mining, visual analysis and computer vision, so as to further improve the competitive strength and performance of outstanding athletes.
The laboratory has obtained a series of research results, and has published more than 20 papers on intelligent collection, visual analysis and prediction of game data such as table tennis, football, basketball and badminton in top international computer science journals.
In the past five years, the laboratory has also undertaken 8 Olympic scientific research projects of the General Administration of Sports, developed the big data platform of the national table tennis team, and made outstanding contributions to the national table tennis team's outstanding results in international competitions such as the World Championships, the Team World Cup and the Tokyo Olympics from 2019 to 2022.
The table tennis big data platform serves the Chinese table tennis team to prepare for the Tokyo Olympics
Digital Science Training Results:
- Intelligent perception of table tennis data
- Data-driven table tennis batting technique training
- Simulation of table tennis techniques and tactics
- National table tennis team big data platform
Achievement 1: Intelligent perception of table tennis data
The perception of sports data is the basis of scientific training. In the table tennis project, a human-machine combination of table tennis data intelligent perception method is required to collect information such as the player's hitting technique, position, and the trajectory, speed, rotation and landing point of the ball.
For this reason, the laboratory proposes an annotation analysis framework for video data annotation and analysis of network motion. The framework integrates computer vision models for scene detection and object tracking, and uses the model output to create a series of anchor points to represent important events in the game.
Ping-pong EventAnchor workflow
By interacting with these anchors, users can quickly find the information they need, analyze relevant events, and ultimately achieve data annotation for simple events and complex player actions. Based on this framework, the table tennis labeling system is realized. The evaluation results show that the system can significantly improve the performance of user data annotation, and its efficiency is twice that of the original method.
The user interface of EventAnchor
Outcome 2: Data-driven table tennis batting technique training
Table tennis is a skill-oriented sport. In training, the player's hitting technique is often the coach's primary concern. In the traditional training mode, the coach can only observe the trajectory of the player's racket swing through his eyes or video, but cannot know the speed and angle of the racket swing. Therefore, there is an urgent need for a table tennis batting training method that combines sensor technology and visual analysis technology.
To this end, the laboratory proposes to optimize the traditional experience-driven training method Tac-Trainer to improve the efficiency and quality of training. Tac-Trainer consists of four parts: device configuration, data interpretation, training optimization, and result visualization.
The device configuration is shown in the figure below, the sensor is placed on the handle of the ping-pong racket and the player's limb to collect the flight speed and rotation speed of the ball. Based on the data collected by the sensor, the athlete's hitting position and hitting technique are further inferred by the data inference model. Subsequently, based on machine learning technology and table tennis technique and tactical theory, a model for quantitatively evaluating the quality of hitting is designed.
Data-driven table tennis training scenario setup
The data and models are finally presented to the user in the form of a visual analysis system. From the system, athletes can discover lower-quality shots as well as recommended ways to optimize.
Visual analysis of table tennis training data and batting optimization system
Achievement 3: Simulation of table tennis techniques and tactics
Based on the non-repeatability of competitive sports competitions, simulation analysis has special significance in competitive sports science research. Simulation analysis can provide coaches with forward-looking recommendations to improve athletes' competitive performance in future competitions. The existing simulation analysis generally adopts the first-order Markov chain model, but this model is not completely in line with the actual situation of table tennis games, such as serving and rushing.
Therefore, in order to complete the simulation analysis of table tennis techniques and tactics more accurately, the research team proposes a simulation analysis algorithm of table tennis techniques and tactics based on the mixed second-order Markov chain model, which extracts the relationship between continuous multi-beat shots in table tennis games to improve the accuracy of simulation analysis.
Table tennis technique and tactical simulation analysis algorithm
Design and implement the visual analysis system Tac-Simur for table tennis game simulation analysis, present the process and results of simulation analysis in an intuitive and easy-to-understand form, provide effective analysis tools for coaches, athletes and game analysts, and explore and discover effective strategies and tactics.
Table tennis simulation visual analysis system
Achievement 4: Big data platform of the national table tennis team
Table tennis is a banner of mainland competitive sports and a traditional advantage of the mainland, but it has also faced severe challenges in recent years. Competitors such as Japan, South Korea, and Germany have significantly improved the competitive level of athletes by introducing scientific and technological means to imitate the playing styles of different opponents. Therefore, we also urgently need to introduce cutting-edge sports technology to ensure the long-term prosperity of the mainland's advantageous projects.
With the support of the Olympic science and technology research project of the General Administration of Sports for four consecutive years, the laboratory has built a table tennis big data platform. The platform follows the idea of human-computer intelligence organic collaboration, through the comprehensive application of artificial intelligence, human-computer interaction, database, cloud computing, visual analysis and other computer technologies, provides a convenient workflow of "collection→ storage→ management→ analysis → application", making analysis more intelligent, decision-making more scientific, and promoting the sublimation of table tennis big data utilization from "available" to "easy to use".
Table tennis big data platform
The table tennis big data platform has been applied to the preparation of the Chinese table tennis team for the Tokyo Olympics. The platform first uses computer vision, visual analysis and human-computer interaction technology to obtain detailed data of game skills and tactics in real time from live video, which are fused with the reverse evaluation results of more than 8,000 existing videos, and through big data simulation and deduction, the winning tactical strategy is discovered, and finally the big data platform deployed in the cloud is pushed to coaches to assist decision-making.
What else is the lab doing?
At present, digital science training has made many attempts in the field of table tennis and achieved rich results, making the Chinese team continue to move towards "faster, higher and stronger". In order to make scientific and technological achievements better serve sports, the laboratory is exploring new fields related to sports, such as basketball situation analysis system and football situation analysis system. These sports systems integrated with science and technology indirectly promote the continuous breakthrough of human athletes' own quality, and I believe that there will be more scientific and technological achievements in all aspects of sports and life in the future, praise for the empowerment of sports by science and technology!
*The pictures in the article are original by the company, please do not reprint without permission
The article "What is this system" section is excerpted from Popular Science China