Graph based visual object tracking

Object tracking is viewed as a two-class ldquoone-versus-restrdquo classification problem, in which the sample distribution of the target is approximately Gaussian while the background samples are often multi-modal. Based on these special properties, we model the visual appearance via graph approach, which is a semi-supervised approach. The topology structure of graph is carefully designed to reflect the properties of the sample´s distribution. The confidence of sample´s label is computed via random walk with restart (RWR). The primary advantage of our algorithm is that it keeps the appearance of object via semi-supervised method. Experimental results demonstrate that, compared with two state of the art methods, the proposed tracking algorithm is more effective, especially in dynamically changing and clutter scenes.