Abstract: How to solve the problem of water and shadow mixing in urban area is important but difficult.However,the difference between the position and shape of the shadow contour at different time provide a potential way to solve this problem.The objective of this study is to propose a method of urban water body information extraction base on the multi-temporal shadow contour difference (WMSD).Two remote sensing images of GF-1 and Landsat 8 were used to test this propose method in Tianhe District,Guangzhou.At the same time,we used some methods of NDWI,MNDWI and SWI to extract water information in this same area,and carried out precision comparative analysis.Study results show that the classification accuracy of WMSD is over 88%,and compared with NDWI,SWI and MNDWI methods,the accuracy of WMSD increase by 8.50%,9.50% and 4.67% respectively,which proves that the WMSD method can solve the problem of the confusion between the shadow and the water extraction information,and can be a feasible method for the extraction of urban water body by remote sensing information.

Key words: GF-1, water extraction, shadow contour, differential image