Abstract: Lane line recognition is the key technology of lane keeping system in automatic driving. In order to improve the accuracy of lane line recognition of large radius circular curve, this algorithm divides the curved lanes on both sides into multiple regions, and uses straight line segment and Bezier curve to fit. Firstly, the image is preprocessed in the region of interest (ROI), and the turn is determined by the distribution of the feature pixels of the lane line according to the change law of the lane line. Then, the region of the left and right lane lines is divided according to different proportions and ways of the turn, and the line segments in the corresponding regions are identified by Hough transform, and the line segments less than the threshold are connected by lines. When the threshold value is higher than the set threshold value, the Bezier curve is used to connect and finally recognize the complete curved lane line. The experimental results show that the algorithm has good robustness to the recognition of large radius circular curve.

Key words: automatic driving, lane line recognition, Hough transform, Bezier curve