A Novel Fast Segmentation Method of Unstructured Roads

In this paper, a method is presented for road surface segmenting in unstructured roads, which have no lane marking. When the original color images are preprocessed using mathematical morphology, binary image is acquired by a threshold in color space. Next, as the deformable template matches in the binary image, some points of road centerlines can be gained. Then, the road surface can be segmented by the points of the road centerline. The method, DTMARF, is more robust in noisy condition, and costs less computing time than other road interpretation techniques. DTMARF has been integrated into a navigation system, which has driven a test vehicle successfully on off-road in many types of weather conditions.