基于半变异函数的多极化SAR图像地表淹没程度分析

洪涝灾害范围的提取能够为灾害的动态监测、评估提取重要的数据。由于半变异函数能够反映图像数据的随机性和结构性,可以很好的体现地物的空间自相关性。而地物在SAR图像上也表现出很好的空间自相关性和纹理特征,各种淹没程度的地物在图像上也表现出不同的空间自相关性,所以将特定窗口下特定步长的半变异函数应用到地表淹没程度分析,证明该方法的可行性。

Semivariogram-based Analysis of the Land Cover Flooding Extent from Multi-Polarized SAR Image

The extraction of flood extension can provide important information for the flood dynamic monitoring and evaluation.Synthetic Aperture Radar(SAR) becomes more and more important in flood monitoring and evaluation,for that its all weather Day/Night capability allows the timely collections of data for flood.And multi polarised SARs can provide more data under different polarized mode than the normal SARs.Semivariogram,as a statistical method developed in Geostatistics,can reflect the randomicity and structure of the image data,so it can reflect the special auto correlation of the land cover very well.Usually,land cover can put up good special auto correlation and texture characterize,and every land cover under different flooding conditions can put up different spatial auto correlation,so the semivariogram under special size moving windows and special lag can be used to the analysis of the flooding extent.Here we tested this method in Poyang Lake using ENVISAT ASAR alternative Polarisation data,which was acquired on October 29th,2004,and included VV/VH polarization mode.The method includes three phases: First,select sample areas and compute the semivariograms,then decide the windows size and lag;Second,compute the whole image's semivariance pixel by pixel,then get a image called "semivariance image";Third,using unsupervised classification method to classify the semivariance image,and by emerging the same flood type to get the flood extent.