中文摘要：

高分辨率遥感道路网络的自动提取在城市信息更新等方面具有非常重要的意义。在综述国内外道路信息提取进展的基础上，本文提出一套基于特征基元的道路网提取方法体系。即采取自下而上的研究路线（“影像像元-特征基元-道路单元-道路网络”）：首先通过影像大尺度的区域划分获取道路区域，在此基础上进行小尺度分割，提取出特征基元；然后根据基元的形态、走向、亮度、纹理等特征对基元进行模式分类，识别出道路单元；最后根据道路网语义规则将道路单元进行形态学处理及拓扑连接，形成道路网络。

英文摘要：

There is very important significance to extract road network automatically from high resolution remote sensing image in updating urban information. This paper proposes a set of techniques scheme for extracting urban road network based on feature units. The processes from clown to upis adopted in the scheme, that is from image to feature units, then to road units and then to road network. Firstly, road region is acquired by segmenting images in a big scale. Then, segmentation in a small scale is executed for feature units. Moreover, road units are identified by classification according to feature of units which includes shape, direction, brightness, texture etc. Lastly, road network is made by morphological and topological operation.