A characteristic model for each target according to the color histogram and kernel function is built.Search range of each target is estimated by using the moving direction in the previous frame. Hellinger distances between the target model and the candidate targets located in the search region are computed

and the candidate target with the smallest distance value is identified as the actual position of target in the current frame. In addition

to solve the occlusion problem frequently arisen in the tracking of multiple objects

a weight strategy for pixels contained in the occluded region is proposed to decrease the influence of these pixels belong to each moving target. The effectiveness and superiority are verified by applying the proposed algorithm to multiple video sequences. The experimental results show that the proposed method outperforms other multi-object tracking method.