中文摘要：

随着城市大比例尺道路网数据库的建设，如何从中派生其他较小比例尺道路网数据库成为重点研究的内容之一。通过道路网的结构特征识别，能够较好地识别其特征目标；Stroke算法能够较好地提取道路网中的主要道路，即能够较好地保留道路网昀框架结构；而极化变换能够较好地对离散目标进行选取。因此，把道路网空间结构识别、Stroke和极化变换等有机结合，首先进行道路网预处理，识别特征道路，然后借助stroke进行主要道路网的识别与选取，最后对剩余较短的离散道路采用极化变换进行选取。实例证明，道路网结构特征识别、Stroke与极化变换的结合，优势互补，对道路网拼行自动综合能达到更好的效果。

英文摘要：

With the development of large scale city road network databases, how to derive other small scale road network databases from them has been one of the urgent research aspects. Structured characteristic identification technique can pick up characteristic features of road network well, and Stroke algorithm can get framework of road network, while polarization transformation algorithm is well known to select scattered point features. So, the structured characteristic identification technique, Stroke algorithm and polarization transformation al- gorithm were integrated together and were made into one new combined algorithm. Example was made and the following advantages oi this new algorithm was given. First, it could pick up characteristic features with structured characteristic identification technique; second, it got framework of road network using Stroke algorithm; finally, the rest scattered road segments were instead with their center points, which would undergo a selection operation with polarization transformation algorithm. Thus, the better generalization results of road network with this combined algorithm was got.