ABSTRACT
The problem of autonomous land vehicle navigation has attracted considerable attention in recent years, e.g., [1-15]. Navigation systems have been developed for road scenes of various complexity using different sensors. Some of these systems incorporate object recognition modules and aim at complete scene un derstanding, e.g., [9]. In many cases attention is primarily focused on determin ing the road edges, which can then be used for autonomous navigation. Various algorithms that employ properties of road geometry [10, 13] and intensity e.g., [3], have been developed for specific road conditions. Due to the low contrast of images and the required processing speed, road edge detection is, in general, very difficult and still remains an open research problem. The difficulty of road edge detection is somewhat eased by more flexible requirements on the accuracy of edge detection because it is necessary to obtain only a rough estimate of the road edges rather than very accurate positions of individual edge pixels.
