Abstract: Based on a high-speed intelligent vehicle with machine vision,an improved algorithm of lane recognition is proposed.In order to obtain fine gradient images, graying,median filter,edge enhancement and optimal threshold are adopted to process images taken by experimental vehicle's CCD camera.Lane edge is detected according to its feature model by an improved Hough transformation.By focusing on area of interest(AOI) in pretreatment and detection,processing time is dramatically reduced.Moreover,the reliability of lane recognition is improved. The experimental result shows that this algorithm is of high efficiency,reliability and robustness.