An auto-surveillance net based on expressway monitor video

With the rapid growth of expressway in China, the monitor video in expressway is also increased, and a surveillance-net is gradually formed, in order to manage all the videos automatically, in this paper, a framework of real-time auto-surveillance net based on a robust recognition and tracking strategy is studied. Recognition is focused on the ROI of the video, and a pixel weight based high-level information abstraction is used to get the detail information of the objects. A hypothesis based tracking is used to maintain correspondences between objects identified at successive recognition instants to provide spatio-temporal trajectories. Results of experiment by dealing with expressway video are obtained, which demonstrate the accuracy and time responses.