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.