Research of flow calculation system based on texture analysis and ROI tracking

Due to affecting by the human body irregular movement or other appendage while detecting human traffic in the channel, the accuracy of flow calculation system is unsatisfactory. We propose a new robust way to detect and count walking persons in channel areas, which mainly includes four aspects: (1) to avoid the human bodies being blocked mutually by vertically fixed monochromatic cameras; (2) to determine the approximate movement region by a simple background subtraction approach; (3) to confirm the pedestrians´ head region by windows´ sliding match in movement regions of continuous frames based on analysis of similar texture, which are as region-of-interests (ROIs); (4) to track ROIs in the next frame images by the ROI tracking algorithm and count ROIs of passing count line. The test shows that the method can effectively solve problems of shadows and the segmenting pedestrians counted in channels, and improve the accuracy of flow calculation system.