The incremental PCA tracking with negative samples

Object tracking is a difficult task in computer vision, which is usually affected by color, surrounding illumination, variation of the object´s appearance and other factors. In previous years, many algorithms can only set up fixed appearance models to track object. Recently, more and more tracking algorithms have been proposed to deal with object appearance variation and illumination change. However, these algorithms are easily influenced by the background and can only track the object for a short time. A novel incremental principal component algorithm with classifier detection is proposed to solve the drifting and long-term tracking problems. Numerous experiments demonstrate that the proposed algorithm is more robust than several state-of-the-art algorithms.