Visual surveillance for moving vehicles

Abstract

An overview is given of a novel vision system for locating, recognising and tracking multiple vehicles, using a single monocular camera mounted on a moving vehicle. 3D model-based techniques are employed to obtain detailed information about vehicle movements. Egomotion estimation is performed by inter-frame tracking of features on the ground plane. Vehicle detection and hypothesis generation is performed using a template correlation technique. Once detected and identified, vehicles are tracked using dynamic filtering. The resulting information about 3D vehicle motions is passed to a collision alert system which can warn the driver of potential hazards. The overall approach is demonstrated on a typical motorway image sequence.