一种基于多重衡量标准共检的车道线检测方法  

A lane detection method based on multi-criteria common inspection

作　　者：王娇 韩加蓬[1] 李煜 马骏 郭栋[1] 谭德荣[1] WANG Jiao;HAN Jiapeng;LI Yu;MA Jun;GUO Dong;TAN Derong(School of Transportation and Vehicle Engineering,Shandong University of Technology,Zibo 255049,China)

机构地区：[1]山东理工大学交通与车辆工程学院,山东淄博255049

出　　处：《现代电子技术》2020年第5期68-73,共6页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(51508315);山东省自然科学基金资助项目(ZR2016EL19)。

摘　　要：针对当前传统车道线检测算法较复杂,适应性及鲁棒性较差,且直-弯车道线检测的准确度较差的情况,提出一种基于多重标准共检的车道线识别方法。对原图像进行灰度化处理后双重设定感兴趣区域(ROI),利用脊度量和Canny检测算子进行边缘特征的提取,双重标准确定直-弯车道线的分界线,利用直-弯车道线模型进行车道线检测。通过实际道路图像对所提方法进行实验验证,车道线检测准确度和拟合程度较高,场景适应能力强,且能够满足实际道路的需要。In view of the complexity,poor adaptability and robustness of traditional lane detection algorithms,and the poor accuracy of straight-curve lane detection,a lane recognition method based on multi-criteria common inspection is proposed.The original image is grayed and double ROI(region of interest)is set.Then the ridge measurement and Canny detection operator are used to extract edge features,and double standard is set to determine the boundary of straight-curve lane.The straight-curve lane model is used for lane detection.The method is verified by experiments on actual road images,which shows that the accuracy and fitting degree of the lane detection are high and the scene adaptability is strong.Therefore,this method can meet the needs of actual roads.

关 键 词：脊度量 Canny检测算子 HOUGH变换 双重标准 分界线 直-弯车道线模型 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]