Dual template algorithm for Mean-Shift template update
均值漂移跟踪的双模板更新算法*

To improve the limitation of Mean Shift lacks method of template update, presented a dual template update algorithm. Firstly, through the comparison of features in target and background, designed the weighting function to weighted the feature distribution respectively. Then introduced a background template and the dual template was constructed based on it. Finally, after comprehensive analysis of similarity degree of candidate object and dual template, the proposed algorithm could accurately determine the tracking status and the cause of interference, and took corresponding template update strategy. Experimental results show that the proposed algorithm can effectively enhance the tracking effect under condition of changeable gesture of target and occlusion, and is more robust.