A method to extract passable area using stereo measurement and projective transformation for small vehicles

Abstract

Road environment recognition is important for safe and easy driving. Various methods which use monocular imaging are proposed for driving assistance on roads such as highways [7]. These methods use road information such as white lines. However, it is difficult to distinguish the passable area from public roads which include obstacles using the 'density pattern' of a monocular image of a complicated road. In this research, we propose a method to extract the passable area using stereo images.
First, road area is presumed using stereo measurement, the principle of epipolar constraint, and Hough transformation. (where, we assumed the road area is plane). This way, the corresponding points that represent the road area in the stereo images are obtained. Second, the parameters of projective transformation are calculated using the corresponding points and transformed into a projective transformation image from the base image (right image). Finally, the passable area is extracted using the difference between the projective transformation image (right image) and the base image (left image).