Adaptive Hybrid Mean Shift and Particle Filter

Abstract

The changing of dynamic models in object tracking can cause high errors in state estimation algorithms. In this paper, we propose a method, adaptive hybrid mean shift and particle filter (AHMSPF), to solve this problem. AHMSPF consists of three stages. First, the mean shift algorithm is employed to search an object candidate near the target state. Then, if this candidate is good enough, it will be used to adapt the particle filter parameters. Finally, the particle filter will estimate the target state based on these new parameters. Experimental results shown that our method has a better performance than the traditional particle filter.