Abstract: Most of the existing water extraction methods are based on multi-spectral remote sensing data, and when they are applied to hyperspectral data, the spectral information is not fully utilized and it is difficult to select appropriate bands, and the shadow effect of buildings in the city is not considered. Therefore, in this study, the modified hyperspectral difference water index (MHDWI) is constructed based on hyperspectral remote sensing image data, using the near-infrared band thresholds and the integral difference in the fixed band range. Four commonly used water indices are selected for experiments and accuracy comparison analysis in Jiaxing, Changsha and Zhoushan areas under different sensors, and the experimental results show that the overall accuracy and Kappa coefficient of the method are 98.50%, 98.76%, 97.35%, and 0.88, 0.85, 0.86, respectively, which are higher than the other four methods. And the misclassification error is significantly reduced compared with the other methods, which shows that the method can effectively suppress building shadows and is applicable to water extraction in different areas with different sensors.

Key words: hyperspectral remote sensing, spectral features, water extraction, building shadows