Object tracking based on support vector dictionary learning

Dictionary learning is widely used to track targets in video sequences. However, a target can be lost during the tracking because of rotation, motion, background clutter, and so on. A dictionary learning method has recently been developed to reduce the chances of missing the target. We developed a new approach using support vector dictionary learning with histograms of sparse codes for a particle filter framework. The representation with support vector can help balance the residual between the candidate and the target. The experiments conducted on challenging sequences demonstrate that the proposed method outperforms seven state-of-the-art algorithms in terms of the overlap rate, center error, and accuracy.