In this paper, an obstacle detection method using binocular road images is proposed and experimental results of the method are reported. A usual binocular method has the problem that it costs much computational time. So this paper presents the method detecting and measuring distance from an obstacle in front by using binocular road images in real time. This algorithm erases information of no use such as road markings and retains useful information from obstacles. Decreasing of feature points provides the reduction of the calculation time. In the experiments on the Tokyo Expressway it was confirmed that obstacle detection was successful at a distance of 130m and processing time was only 0.5 seconds.
Abstract