Robust Visual Tracking Via Local Dictionary Learning With L0 Regularization

Visual tracking is a fundamental task with many applications such as motion analysis,video surveillance and industrial control. In recent years, many works have been proposed, but visual tracking is still a challenging problem due to significant change of target’s appearance caused by heavy occlusion, illumination variation, pose change and cluttered background. In this paper, we establish a robust online tracking system to address this problem in the Particle filter framework.First, we initialize a dictionary using local low-rank features to represent the appearance of the object. In this way, each candidate can be modeled by the sparse linear representation of the learnt dictionary. Second, in order to model the range of appearances, we develop an efficient LO regularized sparse coding model to incrementally learn low-rank features for the tracking target and update the dictionary using the learnt low-rank feature. Finally, we develop numerical algorithms to efficiently solve the resulting non-convex optimization problems. Com-pared with conventional methods, which often directly use corrupted observations to form the dictionary, our low-rank feature based dictionary successfully removes occlusions and exactly represents the intrinsic structure of the object. Furthermore, in contrast to the traditional holis-tic methods, the local strategy contains abundant partial and spatial information, thus enhance the discrimination of our observation model. More importantly, the LO norm based hard sparse coding can successfully reduce the redundant information while preserving the intrinsic low-rank features of the target object, leading to a better appearance subspace updating scheme. To evaluate the performance of our tracking algorithm, we perform our algorithm on challenging image sequences where the target object undergo large changes in heavy occlusion, illumination, pose and scale, fast motion, cluttered background. Experimental results show that our method performs favorably.