ABSTRACT
Target tracking is key problem in the field of computer vision. At present, there are many widely used targets tracking algorithm: Kalman filter [1], Camshift [2,3] algorithm and particle filtering method [1,4]. Among them, Camshift works effectively in simple scenes, but easily lose tracking targets when target scale changes. In addition, many improved tracking algorithm have been proposed, such as [5], solved the half-occlusion problem, but don’t work in full-occlusion and has low robustness when targets scale changes. Such as [6], improved the tracking robustness of scale-changing, but can’t use when occlusion occurs.
