一种高分辨率SAR图像统计建模方法  

A statistical modeling method for high resolution synthetic aperture radar images

作　　者：赵泉华[1] 杜耀西 李玉[1] 王光辉 ZHAO Quanhua;DU Yaoxi;LI Yu;WANG Guanghui(Liaoning Technical University,Fuxin,Liaoning 123000,China;Land Satellite Remote Sensing Application Center,MNR,Beijing 100048,China)

机构地区：[1]辽宁工程技术大学,辽宁阜新123000 [2]自然资源部国土卫星遥感应用中心,北京100048

出　　处：《测绘科学》2022年第3期65-74,共10页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41801233,41801368)。

摘　　要：针对传统合成孔径雷达统计模型不能准确建模高分辨率SAR图像的问题,提出一种基于乘积模型的SAR图像统计模型。在SAR图像乘积模型框架下,假设相干斑服从单位均值Gamma分布,并利用广义Gamma分布精确刻画雷达散射截面积的复杂纹理特性,进而综合上述两种分布构建SAR图像的分布模型。最后,提出基于Mellin变换的分布参数估计方法。定性和定量分析结果表明,相比于经典的统计模型,对于城区、森林、耕地、海面、耕地、灌木和山地等场景,本文模型要优于经典的Gamma分布、K分布和G^(0)分布模型。因此,本文提出的分布模型不仅能够准确描述均质地区的高分辨率SAR图像,而且对非均质地区也展现了其优异的建模能力。According to the fact that the traditional statistical model of synthetic aperture radar(SAR)cannot accurately model high resolution SAR images,this paper presented a SAR image statistical model based on product model.In the framework of SAR image product model,the speckle is assumed to obey the unit mean Gamma Distribution,and the complex texture characteristics of radar cross section(RCS)are accurately described by generalized Gamma distribution(GGD).Then,the SAR image Distribution model is constructed by combining the above two distributions.Finally,a distribution parameter estimation method based on Mellin transform is proposed.The results of qualitative and quantitative analysis show that this model is better than the classical Gamma distribution,K distribution and G^(0) distribution for urban,forest,cultivated land,sea surface,arable land,shrub and mountainous area.Therefore,the distribution model in this paper can not only accurately describe high-resolution SAR images in homogeneous regions,but also demonstrate its excellent modeling ability in heterogeneous regions.

关 键 词：合成孔径雷达 广义Gamma分布 Mellin变换 乘积模型 统计建模 

分 类 号：P237[天文地球—摄影测量与遥感]