单目视觉下结构化车道实时检测

摘要

车道检测是高级驾驶员辅助系统的重要组成部分.本文对视频进行实时处理,实现对结构化车道线的实时检测.首先使用行方向的Sobel算子对处理区域进行边缘增强,接着在处理后的区域使用LSD(Line SegmentDetector)进行线段提取,提取的线段集合包含代表车道线的线段.最后通过线段倾角以及相对位置过滤线段集合,并结合线段稳定帧数来筛选出最佳候选车道线.%Lane detection is an important part of advanced driver assistance systems.In this paper,we processed the video in real-time for achieving real-time detection of structured lane.Firstly we used horizontal Sobel operator for enhancing the edge in the area that we needed to process.And then we used LSD (Line Segment Detector) algorithm to extract lines set,which included the representative of the lane markings.The line angle and relative position were used to filter line set,and combined with the number of stable frame to select the best candidate of lane markings.