车道线图像检测与车辆偏航预警模型构建

Research of Lane Mark Detection and Vehicle Departure Prewarning

研究了车辆偏航预警时道路坐标系与图像坐标系转换精度不稳定问题.对CCD前方道路图像进行了ROI划分,利用自适应动态阈值分割了车道线区域像素,通过Sobel算子提取车道线梯度边缘进行了双重阈值约束,运用Hough变换实现了车道线的直线拟合.根据成像射影原理推导出车道线左右水平倾角与车辆横向偏航率ε之间的关系,给出了基于车道线图像识别的车辆偏航预警策略,建立了车辆偏航预警模型.道路试验表明:当通过车道线图像识别进行车辆偏航预警时,车辆偏航率的平均相对误差为5.7%,满足车辆偏航预警准确性与实时性要求,有效实现了复杂道路环境中车道线识别与偏航预警.

A robust vehicle departure prewarning system,which is a necessary promotion of active safety for high speed vehicle,was proposed based on lane recognition and vehicle lateral motion monitoring.First of all,a self-adaptive threshold method was adopted to segment the pixels of lane mark regions.Gradient edges of lane mark were then extracted by Sobel operator and restrained by dual-threshold method.Next lane marks were modeled as straight lines by Hough transformation.Departure prewarning model was finally built up based on imaging projection theory,from which the relationship between lane mark horizontal obliquity and lateral departure rate ε could be obtained.The proposed method was tested in real road environments.Experimental results demonstrated that the algorithm could recognize lane mark robustly in complex road environments.The average relative error of departure rateε was 5.7%.It is proved that the proposed method could satisfy the precision and real-time requirement of departure prewarning system.