Object robust tracking based an improved adaptive mean-shift method

Abstract

Mean-shift based tracking technique is successfully used in target tracking. However, classic Mean-shift based tracking algorithm uses fixed kernel-bandwidth, which limits the performance when the target's orientation and scale change. In this article, we firstly outlines the basic concepts of Mean Shift Algorithm, and Mean Shift algorithm for target tracking in the visual tracking and its application in visual tracking. Then an improved adaptive kernel-based object tracking is proposed, which extends 2-dimentional mean shift to 4-dimentional, meanwhile combine s multiple scale and orientation theory into tracking algorithm. A multi-kernel method is also brought forward to improve the tracking Accuracy. Finally, experimental results validate that the new algorithm can adapt to the changes of orientation and scale of the target effectively. © 2012 Springer-Verlag GmbH Berlin Heidelberg.