一种自适应动态窗口车道线高速检测方法

A Self-Adaptive Dynamic Window Method for High Speed Lane Detection

为提高结构化道路车道线检测的图像处理速度,提出一种单目视觉自适应动态窗口的高速检测算法.利用栅格法划分出初始图像感兴趣区域,剔除掉与栅格线相交像素点外的其它像素点,进而在保留的像素点中找到车道线特征点,并利用膨胀算法以保留的特征点为基础实时动态产生少量窗口,对这些动态窗口中的图像进行灰度转化、滤波去噪、边缘增强和二值化等处理,得到车道线边界特征,最后利用Hough变换进行车道线拟合.在实车实验中,对实际采集的结构化道路图像的处理速度可达到22 ms/帧左右.结果表明,该算法基本上满足车辆高速行驶时对视觉导航系统的实时性要求.

In order to improve image processing speed for lane detection in structured road, an algorithm with monocular vision self-adapting dynamic window is proposed. Grids are used to mark off the region of interest in the initial image. Then all the pixels are eliminated except those on the intersections of the grid lines, and feature pixels of the lane border in these intersections will be used to generate some dynamic windows using a dilatation algorithm. Then graying, median filtering, edge enhancement, threshold are used to process images in these dynamic windows to obtain lane board feature. At last, lane line can be fitted by Hough transformation. Experiments in structured road showed that the speed of image processing reached about 22 ms/frame and the proposed algorithm could meet real-time request of the high speed vehicle vision navigation system.