Tracking Racket Face in Tennis Serve Motion Using High-Speed Multiple Cameras

Motion capture systems have recently been used in entertainment field, for example, computer graphics, game, movie, and so on. However, it's difficult to measure motion in a natural state in existing motion capture systems, because it's necessary to wear particular equipments such as a lot of markers. This becomes a serious problem in sport motion analysis in actual play. In this paper, we propose a method for marker less motion capture in tennis play by using vision-based techniques. In this method, we track tennis racket face from motion image sequences captured with multiple cameras. We assume that the face of the racket can be modeled as an ellipse, so that we can achieve the tracking by detecting ellipse shape in the multiple view images. We employ high-speed camera system to track the racket because tennis serve motion is too high speed to track it with normal cameras. The multiple view cameras are related with fundamental matrices among the cameras, so that the position of undetected racket in one camera can be found in other positions in other cameras. This enables tracking with good accuracy of racket face. For demonstrating the efficacy of the proposed method, we capture some scene of tennis serve play in a room with multiple cameras, and then the racket can be successfully tracked with the proposed method.