Abstract: Kernel density estimation is first used to establish the color distribution model of a moving target in video sequences to adapt the change of scale.CamShift algorithm is then used to track the object,and its region is located by integrating image moments.To improve the tracking precision,a global update approach is applied to the color distribution model in real time.Experimental results show that the method,which has a strong robustness,can achieve stable tracking when moving object under the conditions of translation,rotation,and partial occlusion,and effectively overcome the impact of scale change.