基于高分辨率影像的道路中心线提取技术研究  

A study of road centerline extraction from high resolution image

作　　者：周绍光[1] 向晶[1] 邱伟[1] 孙金彦[1] 凡莉[1] 

机构地区：[1]河海大学地球科学与工程学院测绘科学与工程系,南京210098

出　　处：《国土资源遥感》2015年第4期21-26,共6页Remote Sensing for Land & Resources

基　　金：国家自然科学基金项目"高分辨率遥感影像中城市道路网的提取方法研究"(编号:41271420/D010702)资助

摘　　要：针对遥感影像道路提取问题,探索了一种基于道路条带自动检测道路中心线的方法。首先基于概率增进树算法获取道路候选点,并通过形态学运算得到光滑和完整的道路条带;然后结合细化运算自动检测道路中心线。但检测的中心线存在"毛刺"且局部曲率变化过大,不符合道路的形状特征。针对该问题,引入了测地距离理论,并用迭代方法去除"毛刺",获得初始道路中心线;再通过Dijkstra最短路径算法优化初始结果;最后根据方向一致性和道路连续性获取最终的道路中心线。采用高分辨率航空影像对上述方法的实验结果证明了该方法的有效性和可行性。In this paper, the strategy to extract accurate road centerlines from acquired road stripe image was explored. The workflow is as follows： Firstly, road candidate points are obtained based on probabilistic boosting tree algorithm, and smooth and integrated road stripes are immediately acquaried by morphology. Secondly, thinning algorithm is introduced to automatically detect road centerlines; nevertheless, the output contained spurs and local curvature of centerlines change much. After that, geodesic distance theory is used to remove spurs. Thirdly, initial results are refined on the basis of Dijkstra algorithm. Lastly, the ultimate road centerlines are obtained according to direction consistency and road continuity. The authors performed an experiment on a high resolution aerial image. The result is satisfactory and shows that the strategy proposed in this paper is an effective method.

关 键 词：道路中心线提取 测地距离 Dijkstra最短路径算法 概率增进树 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]