Particle filter and mean shift are two successful approaches taken in the pursuit of robust tracking.Both of them have their respective strengths and weaknesses.In this paper,we propose an approach in which Kernel mean shift is the dominant tracking method,with a small number of particles being generated to explore further and so resist errors when confidence in the mean shift algorithm is low.Compare with the conventional particle filter,our approach require fewer particles to maintain multiple hypotheses,resulting in low computational cost.