Detection of Obstacle Based on Nocular Vision

A leading obstacles detection and recognition method based on binocular vision is mainly described in this paper. An edge fitting method based on the regional border of fusion segmentation image is used in fitting the edge of road, a viable region and suspected obstacles are obtained through fusing the fusion segmentation image and fitted edge. The method of partial edge contour extraction is used to detection the obstacles; the suspected obstacle edge contour is mapped to corresponding image through the homography, and then the parallax calculation of the two corresponding edges is used to distinguish obstacle and false obstacles accurately. In the outdoor environment, real-time obstacle detection can be realized. Experimental results in real outdoor scene show that the method can effectively detect obstacles.