摘要
目前,高分辨率遥感图象正愈来愈多地应用于土地与水资源的监测和管理,其中,IKONOS图象数据因其具有较高的空间分辨率和人机交互的解译效果,在省市土地利用信息获取及在城市研究中更具有广阔的应用前景.以南京幅的IKONOS图象数据为基础,以城乡接合部的不同试验区为对象,着重讨论了高分辨率遥感图象城市土地利用信息的提取与分类方法 .其主要思想是先应用常用的多光谱分类算法对试验区的土地利用现状进行初始的类别划分 ,然后采用MajorityFiltering技术获取图象的上下文信息 ,并给你根据地物的形状特征和空间关系对初始分类结果进行校正与调整 .研究结果表明 ,当初始的类别划分能满足一定精度时 。
Nowadays, new sets of satellite data with higher spatial resolutions (1～5 m) are continuously used in various land land and water resources monitoring and management aplications all over the world. One of such kinds is IKONOS images which offer a considerable potential for the derivation of information on urban land_use categories with a great visual perception. The present study mainly meant for proposing a method for urban land_use mapping in the fringe area, based on the following approaches such as the standard (per_pixel) multi_spectral classification algorithm to identify the principal land_cover parcels present on the observed scene and, the extraction of contextual information by using ‘majority filtering’ approach to infer land_use from the morphological properties of these parcels and the spatial relationships existing in_between them. During this methodological approach, it was implicated that the initial classification with sufficient accuracy allowed land_use to be inferred from the structural properties and relations. This study was an attempt to apply the proposed approaches by using IKONOS multispectral image with a nominal spatial resolution of 4m for Nanjing city and its fringe area.
出处
《南京大学学报（自然科学版）》
CAS
CSCD
北大核心
2004年第3期362-369,共8页
Journal of Nanjing University（Natural Science）
