Collaborative human tracking with local features in the surveillant camera networks

To track human in a camera network, only the dynamical probability inference is not adequate. In this paper, we propose modified local Speed Up Robust Feature (SURF) to relay track the target person in a camera networks. By adding more interest points and combining local color cue, the computation time of local detecting and matching are greatly saved. Through the experiment results, we see the potential of using the SURF among the camera networks.