Abstract:Model update scheme is a challenging issue to develop a robust object tracking system. This paper puts forward to a collaborative feature learning for accurate tracking with extended particle filter framework. Meanwhile, different model update schemes are employed to select the precise target model. With the collaborative feature learning, the template library with salient features can maintain the rich target appearances. Even when appearance of the target significantly changes, it can still find appropriate appearance model from the template library. Experimental results on several image sequences demonstrate the robustness and accuracy of our approach in a complex scenario.

A Collaborative Feature Learning Method for Accurate and Robust Tracking