Object tracking is a classical problem in computer vision. Existing approaches often use feature descriptors from single cues to build tracking model. This make them fragile under varying illumination conditions and object appearances. Recently, Quang et al have reported an object tracking method that combines different features using graphical model. This method allows weighted contribution from each feature to the final target shift, which permits quick adaptation of the trackers to the environment change and occlusions. The proposed research will extend the object tracking method from Quang et al.