高分辨率SAR相干斑不完全发展的机理、模型及其抑制

With the advent of rapid high resolution techniques, the fully developed speckle of synthetic aperture radar (SAR) has undergone significant change to become non-fully developed. Understanding the speckle mechanisms and building physically based speckle model are crucial for developing the effective despeckling techniques. However, the existing studies mainly focus on the fully developed speckle. Due to that the non-fully developed speckle process is extremely complex, its mechanisms have not yet been fully expounded, and the existing models are empirical-statistical dependent, resulting in a very limited performance of the existing despeckling methods. Based on our previous studies of a kind of non-fully developed and non-stationary speckle and its reduction, this proposal firstly aims to investigate the non-fully developed speckle mechanisms from the perspective of SAR imaging process, which involves many key factors, such as the number of scatterers within a resolution cell and the surface roughness of the observed area. And then the physically based speckle model will be derived. Secondly, a novel non-linear diffusion model is proposed to represent the prior models for high resolution SAR images. Furthermore, the roles of the prior model are explored for preserving textures and edges information in the process of the SAR images filtering. Finally, based on Bayesian inference, the despeckling method is presented for reducing the non-fully developed speckle. The research findings and results of this proposal will enrich the theories of the non-fully developed speckle, and are beneficial in improving the quality of high resolution SAR images.

随着合成孔径雷达(SAR)分辨率的不断提高，其相干斑发生了由完全发展到不完全发展的根本变化。理解相干斑机理、建立物理表征模型是对其有效抑制的重要理论依据，由于相干斑不完全发展过程非常复杂，现有研究主要集中在完全发展方面，对不完全发展的机理尚未阐明，模型大多依赖于经验统计，尚不能完全表征其物理发展特点，因而限制了对其抑制方法的性能。. 本项目将在前期研究的一种不完全发展的非平稳相干斑及其滤波的基础上，基于成像过程系统地研究分辨单元内散射体个数与观测区域非均匀性等诸多关键因素导致相干斑不完全发展的机理，探索具有物理意义的相干斑模型；基于统计学习方法建立面向高分辨率SAR的新型非线性扩散先验模型，分析其在相干斑抑制中对图像纹理保持的规律；最后基于贝叶斯准则，探索不完全发展相干斑的滤波方法。. 本项目的开展将有助于丰富相干斑不完全发展的理论研究，对提升高分辨率SAR图像质量具有积极的参考价值。

随着合成孔径雷达(SAR)分辨率的不断提高，其相干斑发生了由完全发展到不完全发展的根本变化。理解相干斑机理、建立物理表征模型是对其有效抑制的重要理论依据。本项目系统地开展了相干斑的不完全发展研究。主要完成的研究内容为：高分辨率SAR相干斑不完全发展的机理研究，高分辨率SAR系统响应函数旁瓣对相干斑发展的影响机理，高分辨率SAR图像先验模型研究，波长分辨率SAR相干斑模型，以及高分辨率SAR图像相干斑抑制等。特别是，本项目研究发现了当分辨单元内散射体个数小于8个时，相干斑为不完全发展；同时对现有相干斑发展的理论提出了新的补充，研究了系统响应函数旁瓣对相干斑发展的影响；基于滤波器组提出了有效的高分辨率SAR图像先验模型；并在SAR图像滤波领域提出了非平稳相干斑抑制方法，是对现有相干斑抑制方法的丰富完善，使得对高分辨率SAR图像的滤波性能进一步提高。本项目的研究内容和成果丰富了相干斑不完全发展的理论研究，对提升高分辨率SAR图像质量具有积极的参考价值。