Masking Of Shadow And Extraction Of Inland Water Body Using Remote Sensing Imagery

Water body is the main type of land cover,which plays an important part in the LUCC research.Extraction of the water body from remote sensing image is of great significance to the investigation,monitoring and updating of regional water source.Taking the water body around Taihu lake as the research object,this paper proposed a new method for water body extraction using three Landsat/OLI and Landsat/ETM+ images,and validated this method using six ETM+ images covering Taihu lake,Dianchi lake and Qinghai lake.Through analyzing the difference of spectral characteristics between water body and backgrounds,this paper built a complete water body extraction flow,including identification of water extraction index,automatically obtaining of segmentation threshold,quick masking and removal of hillshade and cloud shadow.The results and conclusions are as follows:1.Construction of Water Extraction Index and identification of its thresholdUsing Bswir/BGreen to extract water body,the result had good effects both on large water area and narrow water area with width of 1-3 pixels.The proposed maximum-minimum histogram segmentation method based on LOG filtering can set the threshold,thus can avoid the disadvantages of artificially selected threshold.2.Masking process of hillshade and cloud shadowBy simulating the light conditions of RS image by combination of the shaded relief model and solar elevation angle,the hillshade can be accurately removed.Based on SURF algorithm and the sun azimuth,cloud shadows can be quickly located and masked.3.Comparasion of typical water extraction methodsExtraction results of water body using typical methods vary greatly on smaller water,among which MNDWI method,B component threshold method and humidity component threshold method achieved more abundant details.The main problem is that selecting segmentation threshold artificially is inefficient and the selected threshold is unstable,besides the impact of cloud and cloud shadows on the water extraction is not considered.The ETM+ images in Dianchi area on April 30th,2002,the ETM+ images in Qinghai Lake area on July 31st,2002 and July 12th,2001,the ETM+ images in the Taihu Lake region on January 15th,2001,March 10th,2003 and September 5th,1984 were used for the validation and accuracy evaluation of the proposed method and procedure.The average correct rate is 86.6%,error rate is 1.13%,miss rate is 13.4%.The results show that the method in this paper is of higher accuracy and better adaptability and stability.