Lane detection algorithm based on local feature extraction

An effective local feature extraction algorithm for lane detection is proposed in this paper. First, a lane region of interest (ROI) is determined by the location of road surface appeared in an image. Then, the light intensity and width of lane markings are taken as the local feature. A local threshold segmentation algorithm is utilized to extract lane-marking candidates followed by a morphological operation to obtain the accurate lane. An edge refining procedure is used to eliminate the interference and reduce computational cost. Finally, the lane marking is detected using Hough transform with some subsidiary conditions. With the proposed method, the lane can be accurately detected in conditions of fluctuating and poor illumination, as well as the interference from reflected light can be avoided effectively. The experimental results demonstrate the efficiency of the proposed method.