结构化道路车道标线识别算法研究.docVIP

结构化道路车道标线识别算法研究* 高嵩，秦玲，陈超波 (西安工业大学电子信息工程学院，陕西，西安 710032) 摘 要：针对传统Hough变换算法在无人车结构化道路车道标线识别中出现的中间间断车道标线提取不精准、不完整、有用车道标线信息丢失等问题，提出基于感兴趣附近区域的Hough变换车道标线识别算法。运用传统Hough变换对预处理得到的二值图像进行图像部分分块的车道标线提取，在已提取的车道标线周围划定感兴趣附近区域，找出感兴趣附近区域内的像素与车道标线的辅助度关系，通过辅助度计算累计单元的最大值，得到车道标线的参数。对无人车实时采集的图片进行实验，实验结果表明，该算法在直道、弯道情况下平均处理一帧图片耗时分别为80.42ms、92.22ms，小于课题要求的95ms,具有很好的实时性，且该算法可以有效提取结构化道路的车道标线。 关键词：无人车；车道标线识别；传统Hough变换；感兴趣附近区域；辅助度。 中图分类号：TP391.4 文献标识码：A Structured road lane marking identification algorithm Gao Song, Qin Ling，Chen ChaoBo ( of Electronic and Information Engineering, Xi’an Technological University, Xi’an 710032, China) Abstract：In allusion to the problems in traditional Hough transform algorithm for UGV structured road lane marking identification ,such as extracting the middle of the lane marking inaccurately ,incompletely，and loss the useful lane information, A Hough transform lane marking identification algorithm based on nearby region of interest(ROI) is advanced. In this paper ,delimiting RIO nearby the lane marks that was extracted by traditional Hough transform from the part of binary preprocessing image ,finding out the auxiliary degree relation between pixel in the ROI and lane marking ,calculating the maximum of accumulative unit through this auxiliary degree relation, finally, getting the parameters of lane marks. For the real-time capture images do experiments, the experimental results show that this algorithm processing a frame image respectively need 80.42 ms, 92.22 ms in straight and bending road, less than subject requirements of the 95 ms, has the very good real time, and the algorithm can extract structured road lane marking effectively. Key words: GUV; Lane marking identification; Traditional Hough transform; nearby region of interest, auxiliary degree relation. 无人车在结构化道路上行驶，对道路环境的理解主要是采用图像处理和分析的方法对前方车道线进行Hough变换方法，该方法主要利用传统Hough变换方法建模简单的优点对道路的形态进行建模，利用附近区域特征约束，结合Hough变换空间特征，找出感兴趣附近区域内的像素与车道标线的辅助度关系，从而达到更为全局特征的描述与道路识别的目的。车道线识别的主要研究内容包括两个方面：一方面是对实时采集的图