基于参数化散射模型的三维SAR目标电磁特征提取方法研究

Three-dimensional (3D) synthetic aperture radar (SAR) is an important direction of technology development and research focus for the new generation SAR. Electromagnetic feature extraction is a fundamental problem in SAR imagery interpretation and target recognition. Compared with 2D SAR, the electromagnetic feature extraction for 3D SAR is confronted with more difficulties and challenges, and the problems on this topic in the literatures still need to be resolved. Combined with the advanced development on SAR technology, this proposal aims to explore the intact problem of target electromagnetic feature extraction from 3D SAR. Based on the achievements of our previous work, with the understanding of the mechanism of electromagnetic scattering, and taking the geometry of SAR imaging into account, the relations of target scattering characteristics between bistatic and monostatic of 2D SAR, 2D and 3D SAR will be analyzed, improved and validated in theory more deeply and also some reasonable simplifications and approximations will be made. Then a generalized parametric scattering model which applicable to both monostatic and bistatic 3D SAR under certain circumstances will be constructed. Taking into account that the attributed parameters estimation of 3D SAR parametric scattering model is a complex, high-dimensional, non-convex optimization problem, the corresponding electromagnetic feature extraction method will be presented. Finally, extensive validation, evaluation and analysis will be carried out on the electromagnetic simulated, microwave chamber and outfield measured, and the air-borne (or airship-borne) measured 3D SAR images. It is believed that the research results of this project will greatly promote and accelerate the research and application of SAR especially for 3D SAR target electromagnetic modeling, imaging, feature extraction, imagery interpretation, recognition and so on.

三维SAR是新一代SAR技术发展的重要方向和研究热点之一，电磁特征提取是SAR图像解译与目标识别中的基础性关键问题。与二维相比，三维SAR目标电磁特征提取问题更为复杂和困难，国内外关于该问题的研究还很不成熟。本项目结合先进SAR技术发展方向，瞄准三维SAR目标电磁特征提取这一基础性难题，在项目组前期研究基础上，从目标电磁散射机理出发，结合SAR成像几何，进一步从理论上深入分析、完善和验证双站与单站、三维与二维SAR目标电磁散射特性之间的关系并对其进行合理地简化与近似，创新性地建立在一定条件下对单、双站三维SAR均适用的目标参数化散射模型，并针对三维SAR目标参数化散射模型特征属性参数估计这一高维、非凸的复杂优化问题，提出相应的电磁特征提取方法，最后基于电磁仿真、暗室与外场测量及机载（或飞艇载）飞行试验三维SAR图像数据，完成三维SAR目标参数化散射模型与电磁特征提取方法的实验验证与性能评估及分析。本项目研究对推动三维SAR目标电磁建模、成像、特征提取、图像解译及识别等技术的研究与应用，具有重要的理论与实际意义。

三维SAR是新一代SAR技术发展的重要方向和研究热点之一，电磁特征提取是SAR图像判读解译与目标识别中的基础性关键问题。与二维相比，三维SAR目标电磁特征提取问题更为复杂和困难，国内外关于该问题的研究还很不成熟。本项目结合先进SAR技术发展方向，瞄准三维SAR目标电磁特征提取这一基础性难题，在国内外以及项目组已有研究基础上，对基于参数化散射模型的三维SAR目标电磁特征提取方法进行了研究探索。1）研究了三维SAR目标参数化散射模型的建立，从理论上推导并修正了双站SAR目标三维属性散射中心模型，并通过与FEKO高频电磁计算软件的实验对比，验证了模型的有效性；2）研究了基于参数化散射模型的三维SAR目标电磁特征提取，基于图像域区域解耦合的处理思路，重点针对三维SAR目标电磁特征提取中的参数初始化与参数优化及散射中心结构选择等关键问题，提出了参数初始化和参数优化相结合的参数估计方法，基于多分类器组合的散射中心结构选择方法及基于改进的平滑l0范数（Sl0）的三维SAR图像稀疏重构方法等；3）开发了高分辨SAR目标三维属性散射中心特征提取算法软件，通过对基于FEKO电磁计算软件产生的仿真数据以及美国空军研究实验室(AFRL)发布的“Backhoe Data Dome”数据集的实验结果的性能分析与评估，验证了基于参数化散射模型的三维SAR目标电磁特征提取方法的有效性。在此基础上，本项目还密切结合电磁特征在SAR 图像判读解译与目标检测识别中的实际应用，提出了新的方法。基于本项目研究成果，在国内外核心期刊和会议发表学术论文54篇，其中SCI检索19篇、EI检索20篇，申请国家发明专利9项，获软件著作权7项，培养博士研究生5名，硕士研究生11名。