Robust Tracking Based on Improved Mean-shift and Hybrid Approach

A simple but useful and robust tracking algorithm is proposed. Combining efficiency and reliability is our major objective. The main contribution of this paper is to improve and to combine tracking algorithms to obtain a hybrid tracking method which achieves better tracking effect intuitively, particularly under occlusions. A major disadvantage of mean shift is losing the layout information of targets' colors. Mean shift is modified aiming at this drawback. After analyzing pros and cons of several methods, a hybrid tracking algorithm which relies mainly on the improved mean shift, assisted by Kalman filter and particle filter is proposed. Experiments show that tracking accuracy is effectively improved.