RACKET: Real-time autonomous computation of kinematic elements in tennis

This paper proposes the use of a Visual Sensor Network (VSN) for tracking the motion of tennis players on court in real-time. The proposed autonomous and wireless VSN nodes are miniaturised and powered by battery, making them ideally suited for monitoring training sessions and matches at any location. With the proposed framework, the player is tracked in the image plane using a statistical background model and efficient on-node processing. To improve the usability of the system, the normal markings on the tennis court are used as a calibration grid and the calibration algorithm is implemented with on-node processing. The node further incorporates an HTTP server to simplify transmission and interrogation of the on-node processing results by using mobile devices. The proposed system is capable of tracking a tennis player at 10 to 15 frames per seconds. Multiple nodes are deployed simultaneously either to track several players or to enhance the tracking accuracy of a single player. Features related to motion and game tactics are used to guide the training sessions and refine player tactics.