Abstract: A vehicle distance vision measurement method based on road image key information is proposed, a visual measurement model based on lane plane constraint is established in road image, the slope and vanish point coordinate of road marking line and lane width information collection is finished by boundary constraints Hough transform, automatic calculated height and pitching angle measurement parameters of visual sensor. It selects dual channel Gabor filter to extract 5 scales and 8 direction characteristics samples of the target vehicle, fusion AdaBoost classifier with Cascade screening characteristics effectively, to obtain the coordinates of target feature point parameters accurately. The experimental results show that pitching angle and installation height of visual sensor changes on a small scale have a little influence on measurement results, but the target feature points pixel changes greatly influence measurement results, visual measured values and the measured values of the average absolute error is 1.37 m, the average relative error is 2.38%, average time-consuming 32 ms, compared with traditional method, both of measurement accuracy and real-time are improved, suitable for application in automobile active anti-collision safety system.

Key words: vehicle engineering, vision measurement, lane image information, boundary constraints Hough transform, Gabor filter, AdaBoost classifier

摘要： 提出了一种融合道路图像关键信息的纵向车距视觉测量方法，在道路成像平面内建立了基于车道平面约束的视觉测距模型，运用边界约束Hough变换采集两侧道路标识线的斜率、聚点坐标以及车道宽度信息，自动求解视觉传感器的高度及俯仰角等测距参数。选取双通道Gabor滤波器提取目标车辆的5尺度8方向特征样本，联合AdaBoost分类器与级联Cascade筛选有效特征，快速精确提取目标特征点的坐标参数。实验结果表明，视觉测量值与实测值的绝对误差平均值为1.37 m，相对误差平均值为2.38%，测距平均耗时32 ms，与传统测距方法相比较，测量精度和实时性均得到了提高，适合于汽车主动防撞安全系统中应用。

关键词: 汽车工程, 视觉测距, 道路关键信息, 边界约束Hough变换, Gabor滤波器, AdaBoost分类器