Ball Trajectory and Landing Point Prediction Model Based on EKF Algorithm

This study aims to develop a system for predicting the three-dimensional trajectory and analyzing multiple data points of a ball in sports. This study includes a binocular vision system, an Extended Kalman Filter (EKF) trajectory and landing point prediction system, and an algorithmic data analysis system. The binocular vision system uses two fixed high-speed cameras to record the ball's movement, and three-dimensional coordinates are calculated by triangulation combined with the ball detection function of deep learning. The EKF can solve the Gaussian noise phenomenon caused by image processing and predict the flight trajectory more accurately. The maximum error between the predicted and actual positions of the ball is 0.0165 cm, and the system can also calculate the ball's velocity at each moment. The algorithmic data analysis system developed in this study can analyze the ball trajectory smoothed by post-processing. This system can automatically identify various data during the moving process, such as hitting and bouncing, including the time point, three-dimensional coordinates of the ball at that time point, and which player hit the ball. The system can provide data for professional table tennis athletes as a reference for performance analysis.