Extraction of water body in the west of Taihu Lake based on ALOS image

Quickly and accurately extract the water body information from remote sensing image is an important work in water pollution monitoring and LUCC research. In this paper, water body information of Yixing city in the west of Taihu Lake, China was extracted based on ALOS AVNIR-2 multispectral image, using the multi-level and multi-rule decision tree method. The image was acquired on 29 November 2007, with resolution of ten meters, and was enhanced by linear stretch at first. Then, according to the difference of spectrum characteristic between water body and the other land covers, five extraction indexes including B1, B2, B4, NDVI and NDWI were selected, and dedication rules were identified. Finally, water body information was extracted in the order of non-water body, typical water body and confusion water body. The results show that in the whole area the overall accuracy of this model is 95%, which is 6.04% higher than that of the maximum likelihood supervised classification method when compared with the result of investigation in field and visual interpretation. In the area, the area of water body is 412.78 km², accounting for 20.66% of the total land area.