顾及空间特征的城市主干道提取方法研究

针对高分辨率遥感影像道路提取方法自动化程度低、普适性弱的问题，本文提出了一种顾及空间特征的城市主干道提取方法。该方法首先进行遥感影像滤波，去除道路车辆和影像噪声，采用K—means算法将滤波后的影像聚为2类，即道路类和非道路类；然后进行局部光谱一致性检测，实现备对象分离，依据城市道路的空间分布特征，结合K—means聚类结果，自动识别出道路类；最后，连接细化后中断的道路中心线，得到完整的城市主干道网。利用IKO—NOS、QuiekBird和航空影像对提出的方法进行了试验，结果表明该方法能较好地从遥感影像中提取出城市主干道，且完整性好、精度高、普适性强。

In order to solve the problem of low degree of automation and weak universality- in road extraction from high-resolution remote sensing images, a method of urban main road extraction considering spatial features is put forward. Firstly, the remote sensing images are filtered to remove road vehicles and noises. Then K-means algorithm is utilized to classify the images into two types ： road type and non - road type. Secondly, the localized spectral consistency is detected on the filtered images to separate the objects. Then according to spatial distribution features of urban roads, combined with K-means clustering results, the road type is automatically identified. Finally, the interrupted road center lines are connected and integrated urban main road network is obtained. The proposed methods are tested by KONOS, Quick - Bird and aerial images. The experimental results show that the method can automatically extract the road from remote sensing images with good integrity, high precision and strong universality.