An Algorithm of Mean-Shift Template Update Based On Mixture Gaussian Model

摘要

To improve the limitation of Mean-Shift lack of the template update, an algorithm based on Mixture Gaussian Model is proposed. It treats the target region as “background”, and three Gaussian functions are used to evaluate each pixel value in the target region. After using Mean-Shift algorithm to track the target region in the current frame, we update the Mixture Gaussian Model with the new target region in the current frame, so that the current target template can update automatically with the changing surveillance of selected target. Experiment results show that this algorithm can successful track the changing target surveillance under the condition of change illumination and surface.