基于方向可变Haar特征和双曲线模型的车道线检测方法  

Lane line detection method based on orientation variance Haar feature and hyperbolic model

作　　者：王海[1] 蔡英凤[1] 林国余[2] 张为公[2] 

机构地区：[1]江苏大学汽车与交通工程学院,江苏镇江212013 [2]东南大学仪器科学与工程学院,江苏南京210096

出　　处：《交通运输工程学报》2014年第5期119-126,共8页Journal of Traffic and Transportation Engineering

基　　金："十一五"国家科技支撑计划项目(2009BAG13A04);交通运输部信息化技术研究项目(2013 364 836 900);江苏大学高级专业人才科研启动基金项目(12JDG010)

摘　　要：针对快速路车道线易受多种因素影响而较难检测的问题,提出了一种基于方向可变Haar特征和双曲线模型的分布式车道线检测方法。首先对车载摄像头进行标定,确定图像中车道平面消失线的位置,将车道平面消失线以下部分的下2/3区域作为感兴趣区域Ⅰ。利用边缘分布函数获得感兴趣区域Ⅰ内车道线直线模型倾角,再采用方向可变Haar特征提取边缘特征点并拟合车道线直线模型,利用直线模型参数进一步确定感兴趣区域Ⅱ。提出一种单方向搜索算法,提取边缘特征点并利用双曲线模型拟合获取完整的车道线模型。通过约10 000幅实际道路图片对车道线检测方法进行验证。验证结果表明:检测方法能很好地实现多种环境下的车道线检测,在晴好天气检测率为99.9%,不良天气检测率为99.7%。In order to solve the problem that expressway lane lines were easily affected by many factors, which made them hard to be detected, a distributed lane line detection method based on orientation variance Haar feature and hyperbolic model was proposed. In order to get the disappearing line of lane plane in the image, the camera was calibrated firstly. Then, the lower 2/3 zone below the disappearing line was segmented as the region of interest Ⅰ （ROI-Ⅰ ）. The dip angle of straight line model of lane line in ROI-Ⅰ was obtained by using edge distribution function. Then the feature points of lane line edge were got by using orientation variance Haar feature, and the straight line model of lane line was fitted. The region of interest Ⅱ （ROI-Ⅱ ） was determined by using the parameters of straight line model. A single direction search algorithm was proposed to get edge feature points. Full lane line model was obtained by using hyperbolic model. The lane line detection method was verified by using about 10 000 actual road images. Verification result indicates that lane line detection in a variety of conditions can be achieved well, the detection rate in fair weather condition is 99.9%, and the detection rate in bad weather condition is 99.7%. 2 tabs, 12 figs, 16 refs.

关 键 词：车道线检测 方向可变Haar特征 边缘分布函数 单方向搜索算法 双曲线模型 

分 类 号：U491.223[交通运输工程—交通运输规划与管理]