基于单目视觉的普通道路检测方法

General Road Detection Algorithm Based on Monocular Vision

提出了一种改进的普通道路检测算法,先将单目视觉采集的灰度图像分成左、右2个部分处理,通过在图像坐标内计算倾斜角度和长度过滤可疑边缘;后将左、右图像结合,将图像坐标转换到路面坐标,再根据道路的实际情况再次过滤边缘,提取出最终边界。当前图像检测出的边界成为选择下一帧图像中感兴趣区域的基础,以JJUV-1无人驾驶汽车为实验平台在道路上进行实车实验,验证了该算法的准确率和鲁棒性。

This paper presents a kind of improved general road detection algorithm. It divided the grayscale image captured by monocular vision into 2 parts,by calculating slope of the filter within the image coordinates of suspicious border. transformed the image which was integrated by 2 parts to road coordinate,then based on the road actual situation ,filtrated the border again to extract the final edge,which became the basis to select the ROI in the next image. To verify the algorithm accuracy and robustness,experiments on the JJUV-1unmanned vehicle on campus road was carried out.