Road Edge Detection Based on Vanishing Point Iteration Revaluation

Few traditional road edge detection algorithms can be employed in rural road. An edge detection method based on vanishing point iteration revaluation is presented in this paper. Firstly, the texture orientation of pixel is estimated by Gabor filter, and the corresponding confidence is computed. Secondly, according to the texture orientation, the initial vanishing point is voted by pixels which their confidence are greater than threshold. Thirdly, with the initial vanishing point as starting point, a group of imaginary downward rays are founded, then the orientation consistency ratio weighted by both color difference and double-angle sine function is calculated, and the ray with maximum value is selected as the first road edge. Finally, the optimum vanishing point is given by sampling bilateral edges alternately, and thus two road edges are obtained. A variety of rural road pictures are tested by experiment, and some comparisons are made with the traditional algorithm. The results indicate the method is accurate and robust in rural road.