EARLY OBSTACLE DETECTION USING REGION SEGMENTATION AND MODEL BASED EDGE GROUPING

In this paper, the authors describe an early obstacle detection method which analyzes a sequence of images recorded by a forward-looking television camera that is mounted on a vehicle. Obstacle hypotheses are formulated by evaluating the luminance differences between obstacle and road surface and correlating the motion in succeeding images. A dynamic model-based matching algorithm is used to recognize and validate the obstacle.