基于IPM的复杂环境下道路识别及车辆导航研究

Lane detection of complex environment and navigation based on IPM

在智能车辆自主导航中,对于复杂多变道路的识别和定向,在机器视觉辅助的前提下,往往需要借助许多其他复杂的传感器进行综合定位,增加了车辆的制造成本。造成这一现状的一个重要原因,是由于透视原理所引起的视觉畸变使得我们很难对当前的路面情形作出判断。提出了一种基于色差分割的阈值算法,在此基础上使用IPM(逆透视映射)算法,将相机拍摄到的世界坐标系下的三维景物转换成其在二维平面上的投影图,以此来修复由于透射原理所造成的视觉畸变。

In the autonomous navigation of intelligent vehicles with the aid of machine vision, amounts of sensors are needed for composite locating due to the changeable road conditions, which may lead to extra cost. The most probable reason of this problem is the dififculty in judging the road conditions, which is caused by visual distortion due to the imperfection of the principle of perspective. Thus we proposed a new method of threshold value selection algorithm based on Segmentation of color difference. Then using IPM（Inverse Perspective Mapping） to convert the three-dimension scenes to two-dimension projection, which will ifx the visual distortion to some extent.