A new approach for adaptive background object tracking based on Kalman filter and mean shift

Abstract

Mean shift algorithm is one of popular methods to visual object tracking and has some advantages comparing to other tracking methods. Aiming at the shortcoming of the Mean shift algorithm, this paper proposed a novel object tracking approach using Kalman filter and adaptive background Mean shift. On the one hand, the combination of Kalman filter with Mean shift is suit to handle the case of target appearance drastically changing and occlusion. On the other hand, Bayes law is used to adjust the color probability distribution. It enables objects to be tracked, even when move across regions of background which are the same color as a significant portion of the object. Experimental results demonstrate that this algorithm can track the object accurately in conditions of abrupt shifts, as well as clutter and partial occlusions occurring to the tracking object with good robustness.