A detection and tracking based method for real-time people counting

People counting has wide applications for video surveillance. In this paper, we propose a detection and tracking based counting method. The method involves 2 steps, at the first step, a head and shoulder shape hog feature is extracted to detect pedestrians. At the second step, particle filtering method with color based appearance model, is used to track pedestrians. The method shows great robustness in scenarios of partial occlusions, crowding and background disturbance. With GPU acceleration, our method can be fully real-time and embedded into real applications.