A VISUAL SURVEILLANCE SYSTEM FOR TRAFFIC COLLISION AVOIDANCE CONTROL

Abstract

This study shows how an automatic visual surveillance system can be used to enhance the security and the reliability of a tramway driven by a human operator. We have developed a new approach for extracting the images of moving objects which requires only a sequence of consecutive frames. Gradient operators on current frame and difference picture combined with thresholding and smoothing allows to detect moving edges and to extract moving objects from their background. The smallest surrounding rectangle is assigned to each moving object for tracking purposes. Object tracking provides the data necessary to predict trajectories of vehicles and pedestrians and to predict a possible collision, before actual physical contact occurs. If a risk of collision is discovered by the obstacle recognition system, a collision avoidance procedure has to be activated in order to apply the emergency brake if the driver has not already stopped the tramway.