TM影像的居民地信息提取方法研究

Extracting Residential Areas on the TM Imagery

从对居民地的遥感信息机理分析入手 ,分析了居民地在LandsatTM2、TM3、TM4、TM5、TM7等各个波段上与其它地类的可分性。并探讨了增强居民地与其它地类的方法。同时进一步对各地物的谱间结构特征进行了分析 ,通过分析发现 ,以水泥顶和瓦顶房屋为主的居民地 ,其TM2、TM4、TM7的光谱值比较接近 ,而其它地物除新修道路外 ,都不具备这一特征。这一特征可以用如下表达式来表达 ,TM2 TM4<T1;TM4 TM7<T2 ;TM7 TM2 <T3。通过对T1、T2、T3选取合适的阈值 ,就可以将居民地提取出来。以福清市中部的影像为例 ,通过试验 ,当T1、T2和T3,分别取值为 10、8和 5时 ,即可将该区的居民地提取出来。其点位精度可以达到 82 % ,面积精度可以达到 80 %。其中 ,居民地中的部分绿地没有被提取出来 ,这对灾害的评估更为有效。总之 ,该研究表明 ,通过简单的谱间结构阈值法是可以将以水泥顶房屋和瓦顶房屋为主的集村居民地提取出来。

It is necessary to evaluate the flooded house, population, and property as well as the economic loss. The distribution of house, population, and property are correlated with the distribution of residential areas. Their distributions can be derived from the distribution of residential areas. Therefore, it is very important to quickly obtain the residential area in flood disaster evaluation. Satellite remote sensing provide cheap data resources for delineating residential area. In this paper, the method of extracting residential area from remote sensing data is discussed. FIrstly, the mechanism of remote sensing for residential areas is analyzed. Secondly, the differences of residental area and other landuse type are discussed, and the method of enhancing the differences is explored. Thirdly, the structure feature among spectrum of residental area and the other landuse types is analyzed. The unique feature which can distinguish residental area from the other landuse types is that the TM2, TM4 and TM7 of residental area is very close, especially for residental area mainly consisted of house with cement or tile roof. It can be represented with the formula such as TM2 TM4<T1; TM4 TM7<T2; TM7 TM2<T3. When T1, T2 and T3 are given with suitable value, the residental area can be extracted with the formula. With the example of the center part of Fuqing city in Fujian province, when T1, T2 and T3 are given respectively with 10, 8 and 5, the residential area can be extracted with the position accuracy 82% and the area accuracy 80%. Some green land of residental area is not extracted from remote sensing data, which is benefit to the evaluation of flood disaster evaluation. In conclusion, the residential area can be extracted from Landsat TM with the simple threshold of spectrum structure.