Abstract: According to the characteristics of lane lines, the algorithm of lane line recognition based on multi-stage dynamic programming is proposed. In order to recognize the lane lines accurately, non-uniform B-Spline curve is used to fit the lane lines. The multi-stage dynamic programming is introduced to determine the control points for B-Spline curve. In the algorithm, firstly, according to the region of detection determined by the starting and target points, the candidate control points and the order are obtained after image segmentation. Then, the control points are selected according to the defined cost function and optimization principle. Lastly, the lane lines are fitted so as to realize the recognition thereof. The proposed method is applied to a variety of images, and the experimental results indicate that the algorithm can extract the lane line parameters accurately and effectively for straight line, curved line, continuous line and dotted line, and realize lane recognition stably.

Key words: B-Spline curve fitting, Lane line recognition, Multi-stage dynamic programming, Parallel Mechanism Reconfigurable mechanism Mobility Screw theory