3-D Model Based Visual Traffic Surveillance
Lou Jian · 2003
Visual surveillance in dynamic scenes is an active research topics in computer vision. The aim of visual surveillance is to make it possible that the computer can watch or monitor a scene by automatic localization, tracking and recognition of moving objects and semantic interpretation of their behaviors in the watched scene. This paper aims to realize a task-specific traffic surveillance system which consists of modules for camera calibration, model visualization, pose refinement, tracking, trajectory-based semantic interpretation of vehicle′s behaviors, etc. In this paper, we describe each module to give readers a comprehensive view of a visual surveillance system, and also discuss possible further work.