Flooded Area Classification By High-Resolution SAR Images

Flooding is one of the disasters that widely affected human life.It caused a large number of casualties and property losses every year.Flooded area may cover dozens or even hundreds of square kilometers of area,and flood area rapid extraction has great help and significance for resource allocation,rescue and post-disaster reconstruction.There are great significances in rapid extraction of flood extent and monitoring flood prone area which can help disaster relief and post-disaster reconstruction.As the high-resolution remote sensing satellite data provided for free and widely used,the satellite images can be used in monitoring changes of river channel and flow.Synthetic Aperture Radar(SAR)can provide all-weather condition,wide-range,high-precision images.As there are more and more radar satellites have been launched,it's more convenient for monitoring flood in wide-range areas.Breaching of dikes caused waterlogging in middle of China in 2016,and the disasters has caused significant damage of property.Some latest satellite SAR images in the flood areas were acquired in this paper,researched of water scattering characteristics under different scenarios of multi-source radar satellite images,compared the results of a variety of flood area classification method for application of multi-source remote sensing satellite images to monitor flood areas.The main work and contributions of this paper are:(1)TerraSAR-X,Cosmo-Skymed,Sentinel-1a SAR images were collected,which covered flood areas of Shenzhen city,Wuhan city and Bangladesh country.According to the scattering characteristics of water bodies in the satellite images,the quantitative analysis were done and compared the effectiveness of flood mapping methods.(2)The comparison and evaluation of adaptive median filter,Lee filter,Frost filter and improved Lee filter algorithms were done.A variety of filtering algorithms were used in image preprocessing,then the water scattering characteristics of different images were applied for water extraction,the quantitative evaluation of the extraction results were done.(3)There were effective results in flood monitoring by using Lee filter and improved Lee filter algorithms in the preprocessing,region growing algorithm was applied in water extraction.(4)TerraSAR-X,Cosmo-Skymed images in Wuhu city,Anhui province,in July 2016 were collected,the flood extent was extracted and were compared with real area to evaluate extraction effect.(5)Collected 11 scene Sentinel-la images between 2015 and 2016,which cover the Jamuna bridge area in Bangladesh.Applied the methods and the water characteristics in this study,and combined with high-resolution Landsat-8 images to monitor the change of river region coverage in time series analysis.