基于即插即用框架和二维AMP的稀疏SAR学习成像方法  

Synthetic Aperture Radar Sparse Learning Imaging Method Based on Plug-and-Play Framework and 2D AMP

作　　者：李开明[1,2] 张宏伟 王天润 张强 匡旭斌[6] LI Kaiming;ZHANG Hongwei;WANG Tianrun;ZHANG Qiang;KUANG Xubin(Information and Navigation College,Air Force Engineering University,Xi’an,Shannxi 710077,China;Collaborative Innovation Center of Information Sensing and Understanding,Xi'an,Shannxi 710077,China;School of Equipment Management and Unmanned Aerial Vehicle Engineering,Air Force Engineering University,Xi’an,Shannxi 710077,China;Graduate School,Air Force Engineering University,Xi’an,Shannxi 710051,China;Institute of No.25 of the Second Academy,China Aerospace Science and Industry Corporation Limited,Beijing 100854;Chinese People's Liberation Army 93861 Unit,Xianyang,Shannxi 713800,China)

机构地区：[1]空军工程大学信息与导航学院,陕西西安710077 [2]信息感知技术协同创新中心,陕西西安710077 [3]空军工程大学装备管理与无人机工程学院,陕西西安710051 [4]空军工程大学研究生院,陕西西安710051 [5]中国航天科工二院25所,北京100854 [6]中国人民解放军93861部队,陕西咸阳713800

出　　处：《北京理工大学学报》2025年第2期195-204,共10页Transactions of Beijing Institute of Technology

基　　金：国家自然科学基金资助项目(62371468,61971434,62131020,61871396)。

摘　　要：合成孔径雷达(synthetic aperture radar,SAR)稀疏成像问题主要通过压缩感知(compressed sensing,CS)理论来解决,通过构建正则化优化模型将先验信息引入图像恢复任务.然而,简单的正则化约束难以提供目标复杂的结构信息,尤其是非稀疏场景.提出了一种新颖的基于即插即用(plug-and-play,PnP)框架和深度展开网络(deep unfolding networks,DUN)的二维稀疏SAR学习成像方法.基于线性调频变标算法(chirp-scaling algorithm,CSA)推导出近似观测模型来降低计算成本;使用基于匹配滤波的二维近似消息传递(matched filter-based approximate message-passing,MFAMP)方法迭代求解该稀疏成像问题.为了克服现有稀疏成像方法中先验模型的局限性,在稀疏重建框架中引入PnP先验模型来代替传统的L1范数稀疏正则化器.将成像过程展开为一个DUN,称为基于PnP框架和MFAMP的SAR学习成像网络(PnP-MFAMP-Net).实验结果验证了所提成像方法的鲁棒性和优越性.Constructing a regularization optimization model to introduce prior information into image restoration tasks,it was a main method to solve the sparse imaging problem of synthetic aperture radar(SAR)based on compressed sensing(CS)theory.Due to the simple regularization constraints causing structural information unable provided into the complex target and with non-sparse scenes especially,a novel two-dimensional(2D)sparse SAR learning imaging method was proposed based on the plug-and-play(PnP)framework and deep unfolding networks(DUN).Firstly,an approximate observation model was derived with the chirp-scaling algorithm(CSA)to reduce computational cost.Then the sparse imaging problem was arranged to be solved iteratively based on the matched filter approximate message passing(MFAMP)method.To overcome the limitations of prior models in existing sparse imaging methods,a Plug-and-play(PnP)was introduced in the sparse reconstruction framework to replace the traditional L1 norm sparse regularization apparatus.Finally,the imaging process was unfolded to a DUN named the PnP framework and MFAMP-based SAR learning imaging network(PnPMFAMP-Net).The experimental results show the robustness and superiority of the proposed imaging method.

关 键 词：合成孔径雷达 压缩感知 深度展开网络 稀疏成像 学习成像 

分 类 号：TN957.9[电子电信—信号与信息处理]