Study On Extraction Method Of Inland Surface Water Body Based On Pixel Unmixing

Remote sensing technology has been widely used in flood disaster assessment,water environment monitoring and water resources management.These applications need to study the extraction of water body information,that is,water body extraction is the precondition for remote sensing study.There are many methods to extract the water body information,which mostly based on the pixel that according to the spectral characteristics of the water body.However,no matter how high the resolution of remote sensing imagery,mixed pixels are objective existence,which will inevitably affect the accuracy of water extraction.Therefore,deep into the interior of the pixel and the decomposition of the mixed pixels that not only can improve the accuracy of water body extraction,but also the inevitable trend of quantitative remote sensing development.In this paper,Landsat8 OLI is taken as an example,the different types of water bodies Hongjiannao Lake,Fengjiashan Reservoir,Shahu and Longmen River in the Yellow River Basin are selected as the study areas,which are divided into water body and non-water body.With high resolution images as a reference,the NDWI/MNDWI/MNDWI*threshold method,two-pixel model based on NDWI/MNDWI/MNDWI*and LSMM model that these 7 methods are used to extract the water bodies in each study area,which provide a method for the effective extraction of water bodies in the Yellow River basin with large sediment concentration.The main conclusions are the following:(1)Based on NDWI,MNDWI and MNDWI*,the thresholds were determined by the 0 and OTSU method and also extract different types of water bodies in the Yellow River Basin.The results showed that the MNDWI*threshold method was the most effective,MNDWI threshold method followed,NDWI threshold method is poor,the difference mainly in the water and land boundary and its surrounding small broken water.(2)A new water extraction method is proposed that based on NDWI,MNDWI and MNDWI*,which is based on NDWI two-pixel model,MNDWI two-pixel model and MNDWI*two-pixel model and these three methods are applied to the extraction of different water bodies in the Yellow River Basin.The results show that these three methods can effectively extract the water bodies in the study areas and are sensitive to the small rivers.Besides,based on MNDWI*two-pixel model to extract the best results,based on MNDWI two-pixel model second,based on NDWI two-pixel model is poor.(3)Through improving linear spectral mixing model(LSMM)to extract the different types of water bodies in the Yellow River basin.The results showed that the results of the improved linear spectral mixture model is closer to the high resolution image extraction,especially in the study area where the water is relatively broken,the effect is better.