摘要
在车道偏离预警、防止及车道保持等基于机器视觉的驾驶安全辅助系统中,为解决单一直线模型无法预估道路曲率,而曲线拟合方法计算量大等问题,应用车道线直线模型,提出了一种基于远端直线拟合的弯道曲率估计方法,其特征在于车道线仍使用全局直线描述,估计曲率时对远端候选点进行直线拟合.由于保证了较多的候选点数量且使用Hough变换检测直线,以及利用跟踪、滤波等方法,算法在保证车道线检测鲁棒性的同时,基于远端直线拟合方法实现了对弯道曲率的相对定量估计.采用装有全球定位系统(GPS)的实车试验进行验证,结果表明:所提方法在增加候选点数量以提高曲率估计精度的同时,保障了算法的实时性.
In driver assistance systems of lane departure warning,prevention and lane keeping,to solve the problems of incapable estimating the curvature in traditional linear model and enormous computing work in curve fitting method,an innovative curvature estimation method was proposed based on linear model with global description in lane markings and with line fitting to candidate points of far end.Accor-ding to abundant candidate points with Hough transformation of linear detection,the lane detection and curvature estimation were robustly achieved by the proposed method.The road curvature estimation met-hod was validated by a real vehicle test with GPS.The results show that the proposed method has good curvature estimation accuracy with real time capability.
出处
《江苏大学学报（自然科学版）》
EI
CAS
北大核心
2012年第4期373-378,共6页
Journal of Jiangsu University：Natural Science Edition
基金
国家科技支撑计划项目(2009BAG13A04)
关键词
驾驶辅助
机器视觉
直线模型
曲线拟合
曲率估计
driver assistance systems
machine vision
linear lane model
curve fitting
road curvature estimation
