基于邻域总变分和边缘检测的遥感影像道路提取

针对道路提取的同物异谱和同谱异物难题，提出了一种基于邻域总变分和边缘检测的遥感影像道路提取方法，先对原始影像进行邻域总变分分割，提取出影像中具有均匀材质的地物，并采用Canny边缘检测方法获得原始影像的边缘数据；然后将边缘数据和分割结果融合，使均匀材质区域中的非道路地物和道路分离；最后采用多方向结构进一步提取狭长并具有一定宽度的道路.实验证明本文提出的道路提取方案能同时提取影像中光谱略有差异的道路，能有效剔除和道路黏连在一起的同谱异物地物，并能提取具有一定弯度的道路.

Road extraction from remote sensing images based on neighborhood total variation and edge detection

An approach to extract roads from remote sensing images based on neighborhood total variation and edge detection is proposed. The approach can be used to solve the problem of different objects with same spectrum and same objects with different spectrum in road extraction. Firstly, the origin image is segmented by the method of neighborhood total variation and the homogeneous objects is extracted. Meanwhile the Canny method is used to de- tect the origin image edges. Secondly, the segmentation result and the edge detection results are fused by logical op- eration to separate the homogeneous objects into roads and non-roads. Finally, the method of multi-direction struc- ture is applied to extract the roads result. Experimental results demonstrate that the proposed approach is able to extract roads with slightly different spectrum, eliminate the adhering non-roads and extract slightly curvy roads.