Research on the High Definition Intelligent Surveillance System for Urban Intersection
Shan Huang · Journal of Sichuan University · 2012
A complete set of efficient and robust vehicle detection and tracking algorithms for high definition video was proposed based on the analysis of the complex urban traffic intersections.The intelligent surveillance system achieved integrated traffic surveillance,including traffic flow acquisition,vehicle violation recording and license plate recognition.The efficient two-step vehicle detection method was proposed,which firstly uses the robust inter-frame edge map difference for motion detection,then recognizes passing vehicles by the prominent properties of both the projection of license plate texture and the color of taillights.Vehicle tracking was achieved during a long spatial span by pyramidal Lucas-Kanade method based on large scale FAST corners as robust feature points during daytime and by taillight areas tracking with Kalman filters at night.The system provided promising performance in practice with 96.5% traffic flow accuracy and 87.6% effective ratio of traffic violation recording.