An improved algorithm based on mean-shift for target tracking is proposed in order to overcome the problem of tracking point shift, which comes from partial similarity between target and its background due to lack of color information in the gray images. The contrast mean difference feature, average gradient feature and gray level probability feature of the target are added to the target histogram. The particle filtering is applied to enhance the stability and accuracy for tracking the target in the gray images. The experiment shows that the algorithm can better adapt to the complex background of the moving target, and improves the robustness and accuracy of the algorithm.