Quantitative evaluation of tennis skills by computer vision

Abstract

In this paper, tennis matches are analyzed, and skills of players are evaluated quantitatively by computer vision technology. Two cameras are used to detect the 3-dimensional positions of balls and players. Subtractions of images are used to detect ball images and the prediction of ball positions by using velocity vectors of previous images enables accurate and high-speed ball detection. Players are accurately detected by elimination of white coat lines using image processing technologies. The sequential pictures of tennis experts and beginners were recorded and five scenes, which include tennis play skills, were analyzed quantitatively. The 3-dimensional trajectories of balls and players are calculated and skills between experts and beginners are compared, and the prominent differences between experts and beginners were recognized quantitatively about the trajectories of balls, movement and positions of players, height of balls and speeds of balls.