毫米波全息成像快速反向传播算法研究  

Fast Backpropagation Algorithm for Millimeter-Wave Holographic Imaging

作　　者：牛轶杰[1,2] 柳兴 乔灵博 Niu Yijie;Liu Xing;Qiao Lingbo(Department of Engineering Physics,Tsinghua University,Beijing100084,China;China Academy of Civil Aviation Science and Technology,Beijing100028,China;National Engineering Laboratory for Dangerous Articles and Explosives Detection Technologies,Beijing100084,China)

机构地区：[1]清华大学工程物理系,北京100084 [2]中国民航科学技术研究院,北京100028 [3]危爆物品扫描探测技术国家工程实验室,北京100084

出　　处：《中国激光》2020年第9期300-307,共8页Chinese Journal of Lasers

基　　金：国家自然科学基金委员会与中国民用航空局联合资助项目(U1633202)。

摘　　要：为了在保证图像质量的前提下进一步加快毫米波全息成像的图像重建速度,提出了基于降维策略的快速反向传播重建(DR-BP)算法。基于亚毫米波单站式成像实验(280~320 GHz)以及多发多收正交阵列成像FEKO电磁仿真实验(70~80 GHz)对DR-BP算法进行验证。实验结果表明,DR-BP算法相比仅适用于单站式成像的快速傅里叶变换算法,重建图像边缘干扰少,相比传统的反向传播算法,重建速度大幅提升,本文实验中获得的图像质量相同时,重建速度可提升60倍。To further improve the reconstruction speed under the premise of high-resolution millimeter-wave images, a fast backpropagation algorithm based on the dimensionality reduction strategy(DR-BP) is proposed. The proposed algorithm was experimentally verified via submillimeter-wave single-input and single-output(SISO) imaging(280--320 GHz) and orthogonal array multiple-input and multiple-output(MIMO) FEKO simulation imaging(70--80 GHz). The experimental results indicate that the DR-BP algorithm is able to obtain images with less margin interference compared with the fast Fourier transform-based algorithms for SISO imaging. Moreover, the DR-BP algorithm significantly improves the reconstruction speed—60 times that of the traditional BP algorithm, as shown by the results of the experiments.

关 键 词：全息成像 毫米波 图像重建 反向传播 

分 类 号：O438.1[机械工程—光学工程]