AN ADAPTIVE SEARCH WINDOW ALGORITHM FOR PLAYER TRACKING IN BROADCAST TENNIS VIDEOS

摘要

Player tracking is an essential technique in content analysis of tennis videos. A traditional tracking method consists of defining search window, extracting player's features, and utilizing the features to track in video. Previous methods usually have fixed search windows. Due to the angles of video camera, the court in a tennis video is a trapezoid rather than a rectangle. As a result, the size of the player figure continuously changes while moving from position to the other. Even in the same position, the player body is deformable and very versatile. The fixed search window method not only affects the accuracy of player feature extraction, but also incurs unnecessary calculation. In this paper, we propose an adaptive search window algorithm, which automatically adjusts search window in real-time by referring to player's position and speed. Experimental results demonstrate that the proposed method is effective and efficient for the tracking of players.