Target tracking based on mean shift and improved kalman filtering algorithm
Hongxia Chu, Kejun Wang · 2009
A novel real-time image target tracking algorithm which is based on Mean Shift and improved Kalman filtering algorithm is studied. In the cases of known initial information(position and velocity), measuring point is integrated in tracking window by applying the method of maximum fuzzy entropy Gaussian clustering. The point which has been integrated is inputted to the Kalman filter, and Kalman filter is used to predict the next state's position of the target point. At last, the fast tracking of target is realized by using the combination of Mean Shift algorithm and improved Kalman filter. Result of theory and experiment indicates that the algorithm could keep tracking's real-time performance in condition of image sequences. Accuracy of the target tracking is guaranteed as the target's alternating problem and occlusion problem is improved.