Robust object tracking based on timed motion history image with multi-feature adaptive fusion

Object tracking under complex circumstances remains to be a challenging problem because the appearance of an object can be drastically changed by illumination variations, pose variations and occlusion. This paper proposes an adaptive multiCfeature fusion strategy, in which the target appearance is modeled based on timed motion history image with HSV color histogram feature and edge orientation histogram feature. According to the discrimination measure to the current scene of each feature, the weight of corresponding feature is adaptively adjusted. Experimental results based on challenging video sequences show the robustness and accuracy of our method comparing with several state-of-the-art algorithms.