Visual tracking is a challenging task for object with changing appearance in complex scenes. Online learning and detection trackers are developed to resolve the difficulties such as non-rigid, fast motion, occlusion, rotation and scale change. We propose a novel object based tracking method to achieve robust and accurate object. The integrity of object and structural parts are combined for object tracking. Compressive sensing is employed to represent object as root filter. The sparsity of

more »

... ement matrix is constrained with superpixel segmentation. The part-based model is adopted to filter the local invariant features of parts. The structural constraint strategy between parts and object is developed for adaptive tracking. We test our proposed algorithm on challenging sequences in real world and make qualitative and quantitative analysis. The experimental results demonstrate the method runs in real time and performs well comparable to state-of-the-art tracking.