Combining color histogram and ORB features for robust visual tracking

Robust tracking through long sequences are valuable for many computer vision applications. Illumination and partial occlusion are two main difficulties in real world visual tracking. Existing methods based on hostile appearance information cannot solve these problems effectively. Although there are some local feature based tracking methods devoted to solving these problems, they suffer from the intensive computational burdens, which make them hard to utilize in real scenarios. ORB feature is a brand new local descriptor, which has similar match performances with SIFT while makes much speed improvement. This paper proposes a dynamic tracking approach that combines color histogram and ORB features. The global part of the object model is build up with color histogram and the local parts relies on ORB features. The spatial information of object is also embedded in a simple matching process to increase the robustness of the tracker. The experiments on real world sequences demonstrate that the proposed method can efficiently track objects in challenge circumstances.