一种从SAR图像中提取城市道路网络的方法

提出一种从高分辨率SAR图像中提取城市道路网络的算法.在高分辨率SAR图像中,道路在空间结构上表现为一细长的且宽度基本恒定不变的均匀区域.利用模糊C均值聚类方法对高分辨率SAR图像进行聚类分析,将道路类像素从原始图像中分离出来.为突出道路形状特征,减少冗余信息,对聚类结果进行细化,同时利用跟踪算子消除短线段;以提取道路中心线二值图的像素值作为图像能量,应用Snakes模型检测道路网络.通过实际SAR图像验证,该算法可以准确提取复杂的城市道路网络.

A Method of Road Networks Extraction in Urban Areas Using SAR Images

A method to extract urban road network from high-resolution SAR images is presented.In high-resolution SAR images, the roads are represented as slender regions having almost constant width and reflectance. The road pixels are firstly extracted from the complex background by means of FCM. In order to make the shape of roads clear and to reduce redundant information, a thinning algorithm is applied to the results of clustering, and then a tracing algorithm is used to remove some non-road segments. Due to the speckle noise and shadow of buildings, the central lines obtained are discontinuous. By defining the gray values of the central lines as the image forces, the Snakes algorithm is used to detect the road networks. The experimental results show that the method can extract the complex road networks in urban areas from high-resolution SAR images.