基于OpenCV的车道线检测算法改进研究  

Research on Improvement of Lane Line Detection Algorithm Based on OpenCV

作　　者：罗恒 曹乐乐 金宗毅 李明颖[1] 孟庆宇 Luo Heng;Cao Lele;Jin Zongyi;Li Mingying;Meng Qingyu(Dalian Polytechnic University,Dalian 116000,China;China Railway Construction Electrification Bureau Group Rail Transit Equipment Co.,Ltd.,Changzhou 213000,China)

机构地区：[1]大连工业大学,大连市116000 [2]中铁建电气化局集团轨道交通器材有限公司,常州市213000

出　　处：《北方交通》2023年第2期69-73,共5页Northern Communications

摘　　要：随着自动驾驶技术的快速发展,对车道线检测也提出了更高要求。基于OpenCV函数算法整合,完成了车道线的在线检测,首先通过行车记录仪获得车道线视频的帧图像,对图像进行灰度化、降噪、二值化、形态学等预处理操作,选中ROI区域,再通过面积和外接矩形过滤掉干扰因素,最后进行Canny边缘检测和霍夫直线检测,使用最小二乘法拟合检测到的直线并在原图上画出,实现了在白天、黑夜、有障碍车道环境中的成功检测与准确识别。基于传统图像处理算法进行的该项改进,为自动驾驶领域内的车道线提供了一种更加简单有效的检测途径。With the rapid development of automatic driving technology, higher requirements are also put forward for the lane line detection. Based on the integration of OpenCV function algorithm, the online detection of lane line is completed. First, the frame image of the lane line video is obtained through the automobile data recorder, and the image is preprocessed by graying, noise reduction, binarization, morphology, etc. The ROI area is selected, and then the interference factors are filtered through the area and exterior rectangle. Finally, Canny edge detection and Hoff straight line detection are performed, so as to use the least square method to fit the detected straight line and draw it on the original drawing. The successful detection and accurate recognition are realized in the day, night and obstacle lane environment. A kind of simpler and more effective way to detect the lane line in the field of automatic driving is provided by the improvement based on the traditional image processing algorithm.

关 键 词：自动驾驶技术 车道线检测 OPENCV 外接矩形 最小二乘法 

分 类 号：U495[交通运输工程—交通运输规划与管理]