Unstructured road detection and obstacle recognition algorithm based on machine vision

To realize the unstructured road detection and obstacle recognition, a hybrid algorithm based on minimum error Bayesian decision theory and Hough transform was presented. The Otsu muti-threshold method was introduced into the minimum error Bayesian decision theory and the image was divided into road area and the other area. And then, Hough transform was used to detect the road and off-road borders, and the road area was re-segmented to detect the obstacles on the road. At last, the obstacles were located by outer rectangular bounds. Experimental results showed that the algorithm presented in this paper has good ability to detect the road borders and obstacles. Experiments also indicate that the method was robust when the negative influences such as shadows, changes in illumination, and water stains are slight．