A Method for Water Body Adaptive Extraction from Remote Sensing Imagery Based on Local End Member Spectral Characterization

Precise spectral characterization is fundamental for thematic information extraction. Because different types of water have different spectral features, which result in the difficulty of automatic retrieving water information precisely from the whole film. The present paper explored a new approach to water adaptive extraction based on local end member spectral characterization (LESC). Firstly, through the spectral index calculation the primary water identification was achieved. Secondly, through spatial analysis and automatic end member extraction, we can get the water end member in part of region. Thirdly, according to the end member spectral, we can calculate local end member spectral similarity and histogram of similarity. Finally, through the histogram spectral analysis the optimal segmentation threshold was determined and according to the results the segmentation threshold was adjusted to fulfill water information extracting automatically and accurately. Experiments results show that through local end member spectral characterization the precision of extraction result can be promoted. The proposed method can extract all types of water information precisely and is not affected by different spectral feature.