Abstract: In response to the problem of poor extraction of tiny water bodies in mountainous areas due to different factors, the GF-1 satellite remote sensing image on September 1, 2016 was selected, and three different methods of NDWI, SWI decision tree, and SVM classification were used to were extracted the small water bodies in the Chakou basin of Shanxi province in the gully area, and the extraction effect was visually compared and accuracy verified. The results showed that the high-resolution remote sensing images have better extraction results and higher accuracy,for the small water bodies in mountainous areas than the low- and medium-resolution remote sensing images. GF-1 images can be used to monitor the water body of the river basin; The main factors affecting the water body extraction of the river basin are bright-colored ground objects （mainly buildings） and shadows; NDWI, SWI decision tree, SVM classification method among the three methods, the water body information extracted by the NDWI method is weaker, and the SWI decision tree and SVM classification methods have higher accuracy, but the SWI decision tree method eliminates the influence of buildings and bright-colored ground objects, and clearly distinguishes shadows and water bodies. So it is more suitable for the extraction of tiny water bodies in the watershed.

Key words: extraction of tiny water bodies, GF-1 satellite, high-resolution images, Chakou watershed