Method of pedestrians traffic assessment based on analysis of video data in surveillance systems

This work represent a novel approach for people counting in video surveillance systems. In order to avoid the necessity of objects of interest detection, we propose to evaluate the number of passengers based on visible motion flow. It means that we consider the motion pedestrians not as the motion of individual objects, but as the movement of some mass (flow). This flow is estimated by the methods of background segmentation or by the methods of optical flow. Error estimates of passengers using the technique does not exceed 11%. The accuracy of the proposed method is considerably higher than the accuracy of methods based on tracking objects, especially if there are visual objects overlapping.