Tracking players and a ball in soccer games

In ball games, understanding the movement of players and a ball is essential to analyze matches and tactics. In this paper, we present a system to track soccer players and a ball by using color information from video images. First, our system extracts regions of shirt and pants of each player. It tracks soccer players based on the regions. It copes with the variations of posture and occlusion of players. The ball is too small and fast to extract regions only by color. The system extracts candidate regions for a ball based on both color and time difference image, and tracks the ball using constraint of motion. It maps the position of players and the ball in the image onto that on the field. Experimental results on real image sequences show the effectiveness of this system.