Single object tracking via robust combination of particle filter and sparse representation

The drifting problem is a core problem in single object tracking and attracts many researchers' attention. Unfortunately, traditional methods cannot well solve the drifting problem. In this paper, we propose a tracking method based on the robust combination of particle filter and reverse sparse representation (RC-PFRSR) to reduce the drifting. First, we find the ill-organized coefficients. Second, we propose a diagonal matrix α , whose diagonal line includes each patch contribution factor, to function each patch coefficient value of one candidate obtained by sparse representation. Third, we adaptively discriminate the power of each patch within the current candidate region by an occlusion prediction scheme. Our experimental results on nine challenging video sequences show that our RC-PFRSR method is effective and outperforms six state-of-the-art methods for single object tracking. HighlightsThe first contribution is proposing a suitable contribution factor for each patch to reduce the drifting.The second contribution is giving a reverse sparse representation to make the dictionary more over-complete.The third contribution is using an occlusion prediction scheme to associate each patch with a discriminative power.