Automatic incident detection : slow isolated vehicle and pedestrian on motorway using image processing.

This paper describes an automatic incident detection method using image processing. The algorithms presented are an extension of research undertaken by INRETS. After previous work on stopped-vehicle detection on the motorway, two new types of detection have been developed: slow isolated vehicles and pedestrians. This research has been carried out in cooperation with the Union of Toll Motorways companies (Union des Societes d'Autoroute a Peage) and under ISIS supervision. These functions are integrated in the Tristar-AID system which is now fully exploited by several motorway operators. (A) For the covering abstract see IRRD 881118.