一种基于边缘点投影的车道线快速识别算法

摘要

提出一种基于边缘点投影的车道线快速识别算法,可用于车道线检测和车道跟踪.首先在感兴趣区域内基于投票机制检测消失点,定义目标搜索区域,利用边缘检测信息提取车道特征点.然后沿其梯度方向投影计数,获取车道线上的两个点,再通过置信度判断检测出车道线.最后定义带状区域,实现车道跟踪.采用TMS320DM6437为硬件开发平台,并在集成开发环境CCS下实时仿真调试程序.实车试验结果表明,该算法在DSP硬件平台上运行具有很好的实时性、鲁棒性及准确性,在有阴影、车辆遮挡或雨雾天气等复杂行驶环境下的车道识别和跟踪效果令人满意.%A fast lane marking recognition algorithm based on edge projection is proposed,which can be used in lane marking recognition and lane tracking.Firsdy,the vanishing point in the region of interest is detected based on voting mechanism,object search area is defined and lane feature points are extracted from edge detection information.Then these feature points are projected along gradient direction and counted,with two points on the lane acquired and lane markings detected by confidence judgment.Finally,strip-shaped region is defined and lane tracking is achieved.With TMS320DM6437 as hardware platform,and real-time simulation and debugging program under CCS integrated environment,a real vehicle test is conducted.The results show that the proposed algorithm running on DSP platform has good real-time performance,robustness and accuracy,with satisfactory lane marking recognition and lane tracking effects under complicated environments with shadows,vehicle occlusion,rain and fog.