Superpixel Tracking Based on Sparse Representation

A novel tracking algorithm is proposed that can work robustly in real-world scenarios, in order to overcome the problems associated with severe changes in pose, motion and occlusion. A discriminative model based on the superpixels and a generative model based on global color and gradient features are constructed respectively. Through combining these two models, the distinguishing and invariance of target appearance features description are increased. Furthermore, an update strategy based on sparse principal component analysis is proposed, which can reduce the redundancy of feature dictionary when it updates. A discrimination mechanism is added in the update process of discriminative model to alleviate the drift problem. The experimental results demonstrate that the proposed algorithm performs more stable and robustly compared with several state-of-the-art algorithms when dealing with complex situations such as pose variation, background interference, and occlusion.