高分辨率多光谱遥感影像中城区道路信息的自动提取

An Approach for Automatic Road Extraction in Dense Urban Area from High-resolution Multispectral Satellite Imagery

提出一种从高分辨率遥感影像提取城市区域道路网络的方法。该方法采用改进的数学形态学运算方法对影像进行分割,进而得到粗略道路信息网,然后利用道路网的几何特征实现道路与建筑物的有效区分,最后通过抽骨架的方法获得最终道路网中心线。试验数据为某一城区高分辨率卫星影像,并对最终提取的结果进行了评价,结果表明,所提出的方法能够从高分辨率多波段卫星遥感影像上精确、有效、自动提取城区道路网络。

In this paper, an efficient method of automatic extraction of road network in the dense urban area is proposed. The input data is a satellite image of high resolution in one city. This system is made of four models： data segmentation based on mathematic morphology; discrimination between building and road according to the length and width of object; automatic vectorization by means of skeleton extraction; and finally a model for extraction result evaluation. Experimental results show that this method is efficient in extracting road network in dense urban area form high resolution satellite imagery.