基于滑动ROI的强鲁棒性实时车道侦测算法实现

车道线检测是智能交通系统(Intelligent Transport System,ITS)实现的重要组成部分,文章提出一种在复杂环境中的直弯道检测方法.首先采用带有自适应阈值的横向梯度算子对影像灰度图进行车道线资讯过滤,并采用模糊搜索策略修补过滤后的车道线资讯,提高系统鲁棒性及车道检测的正确率.然后采用滑动ROI(Region Of Interest)窗口策略统计车道边界资讯图梯度方向角,剔除异常梯度方向角的边界噪音资讯,同时采用影像关联方法减少冗余车道线边界资讯,以减少后续Hough转换运算量.直线车道部分提出一种快速查表的Hough转换方法,用于直线侦测;弯道部分引入双曲线对模型,其参数初始值承接于直线车道部分,并通过搜索策略最终确定其k参数,实现弯道部分的侦测.实验表明,此算法在路况复杂的环境中具有较好的鲁棒性,能够实时提供车道资讯信息.

Robust real-time lane detection algorithm based on sliding ROI

Lane detection is an important part of the intelligent transportation system.This paper proposes a method of straight and curve lane detection in complex environment.First of all,the transverse gradient operator with adaptive threshold was utilized to filter the lane information of gray-scale image.Moreover, the filtered lane information was repaired by a fuzzy search strategy,which improved the accuracy of lane detection and system robustness.The sliding ROI（Region Of Interest）strategy,was then used to ac-complish the statistics of abnormal gradient direction angle in the information graph of the lane bounda-ry,meanwhile,image correlation method was utilized to reduce the redundancy of lane boundary infor-mation and the subsequent calculation of Hough conversion method based on rapid tables consulting which was proposed to detect linear lane in the linear lane part.In terms of curve lane part,the hyperbo-la-pair model,whose parameter value was acquired form the linear part and the k parameters were deter-mined through the search strategy,was introduced to detect curve lane.The experiment shows that this algorithm has a preferable robustness in complex road environment and provide real-time lane informa-tion.