Model-based Vision for Road Traffic Understanding


Model-based Vision for Road Traffic Understanding

26th International Symposium on Automotive Technology and Automation (ISATA-93)
Aachen, Germany, September 1993

G D Sullivan, K D Baker.

Department of Computer Science

The University of Reading

email: g.sullivan@reading.ac.uk,

Abstract

This paper gives an overview of a vision system able to locate, recognise and track multiple vehicles in complex traffic scenes, from single or multiple cameras. The system was developed within the Esprit P2152 VIEWS project, and operates in near real-time with quite modest computing requirements. The system makes extensive use of knowledge about the road layout, the camera position and the geometry and expected behaviour of vehicles in the scene. The application knowlege is initially provided by a system engineer, using interactive software tools to tailor the VIEWS system to particular visual tasks. It makes possible a hypothesis-and-test approach to traffic understanding which uses 3D reasoning to classify vehicles and recover their positions and interactions within the traffic scene.


Last Modified: 04:40pm GMT, November 12, 1996