Estimation-correction scheme based articulated object tracking using SIFT features and mean shift algorithm

Object tracking plays an important role in video surveillance system. However, in the field of object tracking, complex object motion and object occlusions still remains challenging topics. This paper proposes a Estimation-Correction (EC) object tracking scheme in real scenarios, combining the strength of scale invariant feature transform (SIFT) and mean shift algorithm. The corresponding SIFT features are used to estimate the position of the target candidate by the scale and space relation between each pair of features. Then mean shift is applied to conduct the local similarity search so as to find a right position and size of estimated candidate with a maximum likelihood. Experiment results demonstrate that the proposed SIFT/mean shift strategy keeps the tracking error in average 8 pixels and improves the tracking performance compared with the traditional SIFT and mean shift algorithm when tracking objects with complex motion and full occlusion.