一种基于单应性变换的车道线检测方法  

A Lane Marking Detection Method Based on Homography Transformation

作　　者：吴琳[1] 邵士媛[1] 李袁[1] 刘春霞[1] WU Lin;SHAO Shiyuan;LI Yuan;LIU Chunxia(Zhengzhou Railway Vocational and Technical College,Zhengzhou 451460,China)

机构地区：[1]郑州铁路职业技术学院,河南郑州451460

出　　处：《郑州铁路职业技术学院学报》2022年第2期21-23,共3页Journal of Zhengzhou Railway Vocational and Technical College

摘　　要：针对辅助驾驶系统中的车道线检测问题提出一种基于单应性变换和直方图分析的方法:首先根据相机参数确定图像中的感兴趣区域,并对图像进行颜色分割和梯度计算以提取候选车道线像素;接着利用单应性变换将图像信息投影到俯视图像,并利用直方图分析提取车道线像素;最后对车道左右两侧提取的标识线像素分别进行二次曲线拟合,从而获得当前车道标识线的几何信息。通过对行车视频图像进行算法检验,结果表明该方法能以30帧/秒的实时速度检测出车道线并计算其几何信息。To solve the essential problem of lane detection in driving assistance system,a fast lane marking detection method is introduced in this paper.At first,a region of interest(ROI)is selected according to the camera parameters,followed by the segmentation by colour thresholding and gradient computation to extractthe possible lane marking pixels.Then the road image is mapped to bird-eye view by homography transformation and histogram analysis is conducted in bird-eyeview forlane marking pixels.Finally,second-order polynomial is used to respectively fit the lane marking lines at the left and right sides,so as to estimate the geometry information of the current lane.The algorithm is tested on driving video images,and the results show that the method can run at real time speed of 30 frame/sec for lane detection and lane geometry information estimation.

关 键 词：车道线检测 单应性变换 直方图分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] U463.6[自动化与计算机技术—计算机科学与技术]