一种弯道标志线启发式分段搜索算法

Curved lane detection algorithm based on piecewise linear model and heuristic search

弯道检测是车辆防碰撞系统的关键技术之一,而基于视觉的弯道识别方法是进行弯道检测的有效途径。为提高弯道识别算法的实时性和鲁棒性,提出一种启发式分段搜索车道标志线的弯道识别算法。结合分段直线模型,采用启发式搜索边界点的算法,在各个动态感兴趣区域(ROI)中搜索车道边界线。对于非连续性车道标志线,结合连续性约束,将检测到的车道线目标拟合为连续平滑的车道线。研究结果表明,该方法能够有效地识别出弯道标志线,识别率可达到86%;识别时间平均达到161 ms/f,能够满足实时性要求。

Curved lane detection is one of important technologies in vehicle collision warning system,and vision-based curved lane recognition method is a useful method.A curved lane detection algorithm combined heuristic search algorithm with piecewise linear model is proposed in this paper.First,the image is divided into several segments,so that the curved lane markings can be approximated by piecewise linear lane model.Next,the lane borderland points are detected by heuristic search in each dynamic ROI.For non-continuous lane marking,combined with the lane continuity constraints,continuous and smooth lane is fitted with detected lane objetcts.The experiment results indicate that the proposed algorithm can effectively recognize the curved line and successful recognition rate is 86%.Besides,the mean recognition time is 161 ms/f which can satisfy the real time request.