Human tracking based on gradient orientation-intensity joint space in infrared image sequences

According to the imaging features of human body in infrared system, a novel human tracking approach in infrared image sequences is proposed. The representing histogram of the human body in a joint space of gradient orientation and pixel intensity is constructed, so as to hurdle the disadvantage of insufficient information when only intensity feature is used. Also, an evaluation criterion is presented to get the optimal bins of the histogram. The above mentioned human representing model is embedded in a particle filter frame, and the human tracking algorithm is designed under this frame. Experimental results under different scenarios demonstrate that the proposed method is more robust and stable than the classical tracking method.