Object tracking using SIFT features in a particle filter

This paper adds sift matching features into the particle filter tracking framework based on color histogram feature, and proposes a dual character tracking algorithm, in which the particle weights are calculated considering both the sift matching features and the color histogram feature. Experimental results show that the algorithm can effectively solve the problem of accumulating errors when inappropriately update the reference template in particle filter simply based on color histogram, especially in cases that the illumination changes or the color feature differences between the background and the target or the targets are relatively small, which enhances the robustness of the system.