Research And Implementation Of Video-Based Pedestrian Counting Algorithms

The technique of pedestrian counting is a focus in the field of computer vision, the number of pedestrians provides every public place important data sources for effective management.In this paper, we gain video image sequences through monocular camera. We complete pedestrian counting by studying the algorithm of moving people detection and tracking in video sequences of a static background.Main content in paper includes background modeling and updating, detection and tracking of moving objects and counting of pedestrian flow. Here we adopt a statistical approach to model the background, and we have three methods to update the background in different situations. We get the human objects by background subtraction. According to the practical application of this paper, we use Mean Shift tracking algorithm combined with Kalman filter to track moving human objects, which could handle the problem objects lost in the case of pedestrians moving too fast or some occlusions.At the end of the paper, there is the flow chart of the pedestrian counting algorithm. And the whole algorithm is described in detail. We test video sequences, and this algorithm is proved to be accurate when detect and track moving people and count pedestrians.