高分辨率卫星影像在城市建筑物识别中的初步应用

Preliminary Application of High-resolution Satellite Image in City Building Classification

在对试验区 IKONOS卫星高分辨率影像预处理的基础上 ,采用 HIS变换法对 1m分辨率全色影像和 4 m分辨率多光谱影像进行了融合处理 ;通过建筑物阴影提取了建筑物高程 ,采用正交小波变换方法提取了图像的纹理信息 ;采用了四种不同的特征图像组合方式 ,对试验区高、中、低层建筑物进行了分类。研究表明 ,由融合图像、高程信息和纹理信息参与的分类结果精度最高 ,对高层建筑物分类的精度可达 80 %。

With the development of remote sensing technique in recent years, high resolution space image shows application potential in many fields, especially in field of city quake calamity evaluation. In this paper, Two IKONOS images of Beijing area are used to classify building types of the city. One is 1m resolution Pan image, the other is 4m resolution and multi spectrum one. After fusion processing, 1m resolution and multi spectrum image is obtained as the basic data for building classification. In order to get best results, some non spectrum information was selected to join the classification process, such as building altitude information which was obtained from the shadow of building and texture. Research shows that different characteristic combination image modes have different classified results. By comparison of four different combination modes, one can see that result, obtained by fusion image, building altitude and texture has best precision, its high building classification precision reaches 80%, and middle and low building classification precisions are all exceed the others results.