改进的图像距离变换方法在车道检测中应用  

Application of an Improved Image Distance Transform in Lane Detection

作　　者：蒋如意[1] 刘洪涛[1] 胡文[1] 王石刚[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《上海交通大学学报》2012年第9期1450-1454,共5页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金资助项目(50875169)

摘　　要：在边缘检测得到二值图像的基础上,提出了一种新的行方向距离变换方法,将其距离图像作为车道检测特征图像,并利用实验验证了行方向距离变换得到的距离图在处理非连续车道线时的有效性及相比于边缘图的优越性.结果表明,行方向距离变换方法不仅能够提供距离信息,而且在处理非连续车道线时能够尽量使其连续化,从而提高车道检测的效率和鲁棒性.Based on the binary image from edge detection, an image processing method was suggested which applies a new ＂row＂ direction distance transform, taking the obtained distance value as the feature for the further lane detection. Moreover, experiments were conducted to validate that the resulting distance map is superior to and more effective than edge map in dealing with discontinuous lane marks. The results show that the new distance transform can not only provide distance information, but also make discontinuous lane boundaries continuous, and so as to improve the efficiency and robustness of lane detection.

关 键 词：车道检测 距离变换 边缘检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]