Object Tracking Based on Local Feature Matching

The feature extraction was analyzed and matching algorithms of object tracking, then we develop an object tracking method based on local feature matching in this paper. First we introduce about the existing feature extraction algorithms and the basic idea of Shape Context. then we propose the tracking algorithm based on local characteristics. in this algorithm, we evaluate the similarity between the two feature points by histogram distance statistics based on χ²-test and we determine the tracking trajectory by votes and prediction to the matching points. It is better matched in target deformation, rotation and occlusion by using this method. the experiment results show that our object matching method is effective.