Research on pattern recognition and sports performance of table tennis game strategy based on big data mining technology

Through in-depth analysis of table tennis players’ technical applications and tactical strategies in the game, in order to explore the common patterns and individual characteristics of the athletes. This kind of in-depth technical and tactical research is of great guiding significance for improving the overall level of athletes and cultivating more excellent players. In this paper, three aspects of improving the YOLOv8 algorithm are proposed: the introduction of an MHSA module, improvement of the loss function, and the addition of a small target detection layer. Using the stereo matching algorithm and parallax method, the 3D coordinates of the target are deduced to reconstruct the trajectory of the table tennis ball movement. Meanwhile, the least squares method is used to fit the scatter coordinates of the trajectory and calculate the trajectory of the table tennis ball. Using the perspective change of the image, the 3D coordinates can be transformed into real-world coordinates of the table tennis table. The improved YOLOv8 algorithm is applied to analyze the table tennis motion trajectory. The spatial motion trajectories of table tennis balls in different situations during the game are different. In match clip 1, the ping pong ball falls from 390.10181mm on the z-axis to -314.0032mm on the x-axis to its lowest point, and then the trajectory begins to rise to a maximum point of 549.2869mm. Analyzing the athletes’ performance in terms of the use of the three types of lane changing strategies, the highest usage rate of the lane changing to forehand position strategy was 37.60%, which was followed by the lane changing to backhand position strategy with a usage rate of 37.43%. The change of line to middle position strategy had the lowest utilization rate of 24.97%.