基于空间概率面的山区居民地遥感信息提取  

Research on Extraction of Residential Area in Mountainous Areas Using Spatial Probability Surface

作　　者：张源[1] 王仰麟[2] 彭建[1] 吴健生[1] 

机构地区：[1]北京大学环境学院,北京100871 [2]北京大学深圳研究生院数字城市与景观生态中心,广东深圳518055

出　　处：《地理与地理信息科学》2006年第4期6-10,共5页Geography and Geo-Information Science

基　　金：国家"973"重点基础研究规划项目(G2000046807);国家自然科学基金项目(40471002)

摘　　要：根据地物之间光谱特征建立的基于知识的遥感居民地信息提取模型是目前居民地信息提取中最普遍的方法,但由于高程差异的影响,其在山区居民地信息提取中效果不理想。以云南省丽江市部分地区为例,在GIS支持下,通过构建多因素空间概率面的方式,综合运用地形和光谱特征信息实现山区居民地遥感信息提取。结果表明,地形差异是影响山区居民地信息提取精度的最主要因素,其影响程度占所有影响因素的50%强;在光谱信息识别的基础上,引入地形这一辅助信息,运用空间概率面能够有效地改善山区居民地信息的提取效果,识别精度从57.5%提高到82.5%。Rapidly and accurately acquiring distribution of residential area is very important for lots of researches such as disaster evaluation, urban expansion and environmental change. The development of remote sensing technology provides a rapid and low-cost way to identify and extract residential areas. Nowadays extraction of residential areas from remote sensing images is rnainly based on spectral analysis. This methodology is quite effective in areas with little differences in altitude such as plains, while it is lost when dealing with areas which distinctly differ in different altitude, such as mountainous areas, because the radiation quantities are observably affected by slope and solar altitudinal angle. Therefore other secondary data needs to be involved to improve precision when identifying residential area in mountainous areas. In this paper,mountainous area of Lijiang City in Yunnan Province is selected to be the study area. Based on GIS, three kinds of spatial probability surfaces concerning spatial probability of residential area distribution in the study area are calculated by spectral information,altitude and slope respectively. And the multifactors spatial probability surface is gained by those three different ones according to Bayes＇ s theorem. Then threshold of probability is corffirmed using total area of residential area in the study area, and the total area in space is distributed according to the threshold, namely, extraction of residential area. Result shows that vast differences in altitude is the most prominent influence on the extraction precision, and precision can be evidently improved using special probability surface, which is 82.5% and higher to 57.5 % only using spectral information.

关 键 词：空间概率面 山区居民地 遥感 地理信息系统 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置] P208[自动化与计算机技术—控制科学与工程]