Object Tracking with Sparse Representation and Annealed Particle Filter

Abstract

In this paper, we propose a new visual tracking algorithm, SRAPF, for object tracking, which is based on sparse representation and annealed particle filter. To find the tracking target at a new frame, each target candidate is sparsely represented by target templates and trivial templates. The sparsity is achieved by solving a l1-regularized least squares problem. After that, Instead of tracking objects in the common particle filter framework, we solve the sparse representation problem in an annealed particle filter framework. Then the candidate with the largest likelihood is taken as the tracking target. In the APF framework, the sampling covariance and annealing factor items are incorporated into the tracking process. The annealing strategy can achieve "Smart sampling" to avoid generating invalid particles corresponding to impossible target object. Both qualitative and quantitative evaluations on challenging image sequences demonstrate that the proposed tracking algorithm performs better in comparison with the L1 tracking algorithm.