知识辅助的对象化高分辨率极化InSAR层析参数反演新理论与新方法研究

The inevitable trend of radar system development is high resolution image and multiple polarimetry. The characters of ground scattering and phase statistics in high resolution images are different from those in median or low resolution images. Therefore, it is necessary to develop new theory and method for high resolution processing and detailed parameter inversion of ground scenes. According to the new problems of high resolution polarimetric synthetic aperture radar (PolInSAR) tomography, such as the enhanced heterogeneity of observed scenes, the deficiency inversion samples of non-dense distributed vegetations, we will give a deeply investigation for the new principles of knowledge-aided and object-oriented PolInSAR tomography. By introducing the auxiliary knowledge and dividing the complicated scenes into different objects, we will give focus on the knowledge- aided baseline optimization and optimal data set obtainment, the adaptive coregistration of multi-dimension signals for high resolution PolInSAR tomography, the object-oriented sample selection and the parameter inversion model for the vegetables of complicated vertical structure, which will give a technical support for the detailed parameter inversion with complicated vegetables’ distribution. Through the investigation of this project, the new model and method of knowledge-aided and object-oriented PolInSAR tomography will be founded. The ability to descript complicated observed scenes of high resolution radar systems will be improved, and the adaption ability of signal processing algorithms will also be improved. Furthermore, the research results will also give a technical support for the existing PolInSAR systems.

高分辨率、多极化是雷达系统发展的必然趋势，地物散射、相位分布等特性在高分辨率时表现出不同于中低分辨率图像的特征，需发展新的理论和方法以适应高分辨率处理及对观测场景精细化反演的需求。针对高分辨率极化InSAR层析处理面临的场景非均匀性增强、非密布植被反演样本不足等新问题，引入辅助知识和对复杂观测场景的对象化处理，开展知识辅助的对象化极化InSAR层析参数反演方法的研究。突破知识辅助的基线优化设计与数据获取方法、高分辨率极化InSAR层析多维信号的自适应配准、对象化样本选择与复杂垂直结构参数反演模型建立等关键技术，为高分辨率极化InSAR层析处理获取场景的精细化信息提供理论和技术支撑。通过研究，预期建立知识辅助的对象化极化InSAR层析处理的新模型与方法，可提升高分辨率雷达系统对复杂观测场景的精细化表征能力和信号处理算法适应能力，并为挖掘现有系统的性能提供技术支撑。

根据高分辨率/超高分辨率极化InSAR 层析处理的新特点和新问题，开展地面场景全方位、精细化的信息获取技术研究。主要进展包括如下方面：(1) 针对高分辨极化InSAR 层析处理面临的场景非均匀性增强，非密布植被/稀疏分布植被反演样本选取，提出了基于超像素处理的样本选择方法解决场景非均匀带来的样本统计特性不一致，同时提出基于归一化相位广义内积的稀疏植被参数反演样本选择方法，解决了植被稀疏分布时的样本缺失问题；(2) 针对复杂垂直结构分布的场景建模与反演，提出了基于DEM辅助的干涉图像配准方法、DEM辅助的干涉基线误差估计方法，有效提升了极化InSAR图像序列的配准精度；并进一步在配准误差的基础上，提出了基于广义散射矢量的稳健极化干涉最优相位图估计方法，有效保证了后续PolInSAR参数反演的相干性信息；(3) 建立知识辅助的极化InSAR 层析反演新模型与新方法，提出了基于最大化相位差最优化方法分离的方法，有效分离出电磁波在植被上、下两层散射区域的干涉相位，能有效消除由于穿透深度引起的植被参数参演欠估计的情况，并进一步根据多航过高分辨机载数据，实现了基于Tikhonov正则化处理的极化InSAR层析反演模型。.通过本项目的研究，建立了知识辅助的对象化极化 InSAR 层析处理的新模型与方法，突破高分辨率极化InSAR 层析的关键技术，提升了现有系统观测性能。