Crowd Segmentation Using SimpleFlow

Abstract

This research is aimed to realize a high accurate crowd segmentation. Currently video surveillance systems are widely used at various places. With the rapid development of computer vision technologies, automatic identification and tracking techniques have been applied to surveillance system. For the crowd segmentation, e not only use the optical flow but also use streakflow. The reason is that the streakflow information is better than the optical flow. We choose Simple Flow which is a non-iterative optical flow technique and running times increase sub-linearly with the number of pixels to denoise and analyze crowd motion accurately. The accuracy of segmentation will be shown where the streaklflow is higher than that of only the optical flow segmentation by some experimental results.