摘要
针对结构化道路上存在非车道线标记干扰的情况,提出一种基于成像模型的线扫描车道线检测及跟踪方法。检测算法中首先对路面图像进行形态学高帽变换预处理,然后建立前方道路图像的成像模型,将图像坐标系中车道参数和世界坐标系中实际车道参数对应,对图像进行初扫描,利用边缘贡献函数及RANSAC算法选取最确定线后,以此线为标准进行二次扫描,得到边缘点后统计边缘贡献函数局部最大值并拟合成直线车道线。跟踪算法中运用Kalman滤波器预测车道线区域,并提取符合标准的控制点拟合成模型为B样条的车道线。试验结果表明:该方法能够快速准确地在复杂环境中提取多个车道线,尤其对存在非车道线道路标记干扰的情况有显著效果。
Considering the interruption of non-lane road markings,authors presented a lane detection and following algorithm on the basis of imaging model.In detection stage,lane image was firstly preprocessed by a morphological top-hat transform,imaging model was created for building relationship of lane parameters of the image coordinate and the world geodesic system（WGS） coordinate.Then the image was scanned for the first time on the basis of imaging model,after that,the edge distribution function（EDF） and RANSAC algorithm were used for the most sure lane detection.Based on the imaging model and the most sure lane,the image was scanned for the second time,local maximum values of EDF which were fitted to straight lane were recorded after the edge points.In following stage,Kalman filter was used for searching lane area,then B-spline model was applied to the lane fitting.Test results show that the proposed algorithm can efficiently and accurately detect and follow multiple lanes in the complex environments,particularly in the presence of interruption of non-lane road markings.
出处
《中国公路学报》
EI
CAS
CSCD
北大核心
2011年第6期96-102,共7页
China Journal of Highway and Transport
基金
国家自然科学基金创新研究群体基金项目(40721001)
教育部高等学校博士学科点专项科研基金项目(20070486001)
