A Robust Tracking Method Based on Particle Filter via Multi-Cues

摘要

In view of the issues of multi-cues selection and fusion in the object tracking, a multi-cues fusion tracking method based on the particle filter is proposed in this paper. This method utilizes the combination between the grey histogram and the SVD (Singular Value Decomposition) invariance as the criteria of the similarity of each particle to enhance the stability. A feedback mechanism is also established according to the-contribution each cue to the final result so that the parameters can be adaptively modified in the process of multi-cues fusion, which helps to guarantee the reliability. Experimental results show the proposed method is superior to the method using particle filter based on the common multi-cues fusion.