Secondly, command EXPAND of ARC/INFO can be used to expand base water bodies for two pixel. It is used to cut out the primary flood from SAR image. and then some of the omitted flood (figure 8) can be got, which is added to the flood extent without shade. The procedure is shown as following (figure 9):


Figure 8 the water bodies including some of the omitted flood


Figure 9 Restoring some of the omitted flood

3.6 Adding the omitted water bodies to the flood extent without shade
Because the resolution of RADARSAT SAR image is 50 meter, It is difficult to identify small water bodies. It can be solved with TM images. Moreover, in some case, water surface become rough owing to wind, which show higher gray level than the normal one, and make some water bodies omitted. In order to compensate the omission, the base water bodies from TM is added to the flood without shade, and the final flood extent is got (figure 10).


Figure 10 The final flood extent

4 Result and discussion
The result is analyzed through visual interpretation of flood in RADARSAT SAR image and water and topology in LANDSAT TM image. The accuracy is higher for plain place, and the mislabeling is very low. Omission is also lower. It is estimated that the accuracy can reach to 90% in plain area. In mountain area, the proposed method of automatically extracting water from RADARSAT SWA SAR is also proved to be effective by Visual interpreting shade and topography from TM and flood from RADARSAT SWA SAR. The accuracy can reach to 85%. More research need be done especially for other SAR imaging mode.

References

• Zhang jiqun. Remote sensing analysis of flood disaster neural network, PH.D. thesis of the university of Wuhan water and hydoelectrics 1995: page s2~3:84~99.
• Dai Changda, Tang Linili, Chen Gang, The theory and practice of applying Landsat TM image to monitor and evaluate forest fire, pest insect and flood, Remote Sensing of Environment, 1993, 8(2): 102-110.
• Sheng Yong Wei, Xiao Qianguang and Chen Weiying, Study of applying FY-1B meteorological satellite to monitor the flood of 1971. Remote Sensing of Environement 1994, 9(3): 228-233.
• Zhou Chenghu, Du Yunyan, Method and model for automatically extracting flood extent from NOAA AVHRR image on basis of knowledge. Journal of natural disaster, 1997, 6(4):100~108.
• YangCunjian WeiYiming Chen Dequing Investigation on extracting the flood inundated area from JERS-1, flood inundated area from JERS-1, Journal of natural disaster, 1998, 7(3):43~50.
• Guo Huadong, Xu Guanhu, Study on application of satellite radar, China science and technology press, 1996,3:203-208.
• Jian Yonghong, Li Diren. The comparison of application four HIS conversion in fusion of SAR and TM image, Journal of remote sensing, 1998, 2(3) :103 _106.
• Fang Hongliang, Yang Xiaomei, Du Yungan, Application of principle conversion and inconversion in fusion of ADEOS AVNIR SX and PAN, Remote sensing technology and application, 1998 13(3).