基于面向对象的城市地物信息提取方法研究

20世纪90年代以来，高空间分辨率遥感影像数据的处理已成为遥感领域中的热点与难点。利用具有人的思维特点的面向对象的信息提取技术，对高分辨率遥感影像中的城市用地进行分类，分析和利用高分辨率影像的空间信息、结构信息与光谱特征等，总结了面向对象解译方法的5个步骤，即影像分割、分类方法的选择、地物种类分类，知识库构建、计算机自动分类。分类结果表明：（1）克服了“椒盐现象”；（2）信息提取的总体精度为92.19％，而且各类地物信息的提取精度均有所提高，特别是利用前期分类的拓扑关系有效提取了城市水体与建筑物阴影。

STUDY ON EXTRACTION TECHNIQUES OF URBAN LAND-USE INFORMATION BASED ON OBJECT-ORIENTED

The processing of high spatial resolution remote sensing （HSRRS） date has become a major focus in the remote sensing field since the 1999s. Object-oriented information extraction technology with human thought was used to extract urban land-use from high resolution remote sensing images. Spatial information, structure features, spectral features and so on of high resolution image were made full use of. The processes that there were five steps （ image segment, the choices of classification methods, classifications of urban use-land, construct repository, computer automatic classification） were concluded by based on object-oriented method. The conclusion were （1）“pepper and salt” noises were discarded. （2） 92.19% overall accuracy was achieved. Furthermore, precision of each kind of object information was also improved. At the same time, water bodies and shadows were availably extracted by rules of topology.