Fragments based tracking with adaptive multi-cue integration

Lichuan Gu, Jianxiao Liu, Chengji Wang

COMPUTER MODELLING & NEW TECHNOLOGIES 2014 18(2) 126-134

School of Computer and Information, Anhui Agriculture University, Hefei Anhui, No.130 West Changjiang Road China, 230036

In this paper, we address the issue of part-based tracking by proposing a new fragments-based tracker with multi-cue integration. First the target template is divided into multiple fragments to get the contribution of each fragment to the possible positions of the target in the current frame, similarity measure is used in edge histogram and HSV histogram in every fragment, and the weights of cues integration are computed adaptively. Second we present a fragment template update mechanism with the discrimination between occlusions and appearance changes. The template is unchanged when the target is occluded and some fragments of template are updated in the case of appearance changes. In the experiments we use the indoor and outdoor test videos which contain the illumination changes, occlusions, and the appearance changes of targets. The experimental results show that our approach has strong robustness and high tracking accuracy.