Abstract: The typical surface feature of Huanghe River estuary wetland is complex and diverse. In this study, a new classification model for coast wetland remote image was constructed using the linear spectral mixture analysis model, combined with normalized difference vegetation index(NDVI) and normalized difference water index(NDWI). Based on CHRIS hyperspectral image, a classification test of Huanghe River estuary wetland was carried, which consisted of 6 kinds of typical objects: phragmites, tamarix chinesis, suaeda, spartina, tidal flat and water, The results show that the overall accuracy of the combined model is 77.33%, and Kappa coefficient is 0.71, increasing 1.6% and 0.02 respectively compared with that from the classical MLC method, and especially, a better classification accuracy is obtained obviously for phragmites, suaeda, spartina and tidal flat.

Key words: Huanghe River estuary wetland, hyperspectral remote image, linear spectral mixture analysis, classification of typical surface feature, NDVI, NDWI