An adaptive tracking window based on mean-shift target tracking algorithm

As the traditional Mean-Shift tracking algorithm fix the size and orientation of tracking window, it can not effectively track the target when the scale and orientation of tracking target has a distinct change. In response to this problem, the proposed algorithm which combines target contour and the principal components of the variance matrix with Mean-Shift tracking algorithm. It uses the target contour Information of the real-time to get the current size of the target and uses the principal components to compute the orientation of the tracking target. Experimental results show that the improved algorithm has a good adaptability when the scale and orientation of the target change.