Tracking Ball in Soccer Game Video using Extended Kalman Filter

The detection of the ball is the first step for tracking in soccer broadcasted video. In some cases, it is difficult to detect the ball by shape and color. Especially, when it overlaps with other objects (line or players). Therefore, we have been suggested a new technique of real-time ball tracking. First, reducing the rate of a missing ball through determining the candidate position of balls rather than attempting to identify the position of ball, computing the distance between the ball and candidate balls to delete the false candidate position of the balls by the threshold. At last, estimating the ball position via Extended Kalman filter. The proposed work has achieved higher accuracy and speed than other methods which are used Kalman filter and template matching or used only a Kalman filter for tracking the ball.