The Research Of Tracking Moving Objects Based On Mean Shift

The efficient tracking of moving objects in complex environments is a challenging task in computer vision and has a wide variety of applications such as video conference, robotics navigation and virtual reality. The accuracy and stability of the tracking system is directly influenced by the performance of the tracking algorithms. Most of these approaches which are mainly feature-based or motion-based are too much complicated and occupy a large percentage of system resources. In this paper, a new method for real-time tracking is proposed, of which the central computational module is based on the mean shift procedure.In this paper, we summarize several classic detecting and tracking algorithms used both under static background or moving background. Both advantages and disadvantages of these algorithms as well as their adaptability to different situations are also discussed.Nonparametric density estimation on which mean shift iteration is based is formulized thoroughly. The derivation of mean shift procedure as well as the strict proof of its convergence is given.This paper presents the approach to the real-time tracking of moving objects based on mean shift iterations. A new tracking approach, which combines the centroid-based tracking and mean shift procedure, is discussed and simulated. The tracking results show that it can effectively reduce the computation and have superior tracking performance.