A Reinforced Road Detection Method in Complicated Environment

Road detection is the most fundamental part of autonomous vehicles. Noises caused by shadows and vehicles on the road have a great negative impact on the road detection. Our method improves the performance under the noisy environment by taking advantage of color information to determine the road curvature. A novel geometrical method is proposed in this paper to select the most matched curvature. A graph-based image segmentation algorithm is introduced to extract the road area which is not affected by shadows and vehicles on the road. A bilateral filter is presented to smooth images while preserving edges and to help detect continuous and perceptually clear road borders.