基于单目视觉传感器的车距测量与误差分析

Vehicle distance measurement and its error analysis based on monocular vision sensor

为了解决结构化道路上基于单目视觉的运动车辆车距测量问题,从针孔模型下摄像机成像的基本原理出发,推导出基于图像中车道线消失点的车距计算公式。车距测量结果只与图像中的近视场点到摄像机的实际距离有关,而无需对所有的摄像机参数进行标定。分析了车距测量中的误差因素,并在前方道路上设置已知距离的横向标线,完成了摄像机不同安装高度、俯仰角及方向角情况下的标线距离测量实验。通过与实际距离比较发现,摄像机安装高度与方向角的微小变化对车距测量的影响可以忽略,而摄像机俯仰角的变化将引起较大的车距测量误差。最后,完成了不同距离处前方车辆车距测量实验,实验结果表明:采用上述方法的车距测量相对误差小于3%,具备了较高的检测精度。

To resolve the problem of vehicle distance measurement based on monocular vision on structural road, the formula of distance calculation based on vanishing point of lane lines is deduced from the basic theory of imaging by a pinhole model camera. The calculation is related only to the actual distance between the camera and the point of near field of view without calibrating all parameters of the camera. Then the error factors are analyzed and experiments for the measurement of distance between the camera and the horizontal markings on the road is performed under different camera installing height, pitching and orientation angles. Compared with the actual distance ,the results indicate that the effect of the small changes of installing height and orientation angles on distance measurement can be ignored, while the change of pitching angle leads to a large error. At last, the experiment of vehicle distance measurement is also performed with a preceding vehicle on the road. The result demonstrates that the relative error is less than 3 % and a high precision is reached.