多块合并压缩感知实时成像  

Multiblock compressed sensing imaging in real time

作　　者：李虎 刘雪峰 姚旭日 刘璠 窦申成 胡钛 翟光杰 LI Hu;LIU Xue-Feng;YAO Xu-Ri;LIU Fan;DOU Shen-Cheng;HU Tai;ZHAI Guang-Jie(Key Laboratory of Electronics and Information Technology for Space System,National Space Science Center,Chinese Academy of Sciences,Beijing 100190,China;Center for Quantum Information Sciences and Key Laboratory of Advanced Optoelectronic Quantum Architecture and Measurements(MOE),School of Physics,Beijing Institute of Technology,Beijing 100081,China;Laboratory of Satellite Mission Operation,National Space Science Center,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China;Beijing Academy of Quantum Information Sciences,Beijing 100081,China)

机构地区：[1]中国科学院国家空间科学中心复杂航天器系统电子信息技术重点实验室,北京100190 [2]北京理工大学物理学院量子技术研究中心和先进光电量子结构设计与测量教育部重点实验室,北京100081 [3]中国科学院国家空间科学中心空间科学卫星运控部,北京100190 [4]中国科学院大学,北京100049 [5]北京量子信息研究院,北京100081

出　　处：《红外与毫米波学报》2023年第1期61-71,共11页Journal of Infrared and Millimeter Waves

基　　金：National Key Research and Development Program of China(2018YFB0504302);the Youth Innovation Promotion Association of the Chinese Academy of Sciences(2019154)。

摘　　要：中长波红外成像探测器成本高昂,成为该波段高分辨成像和实时显示的巨大挑战。本文提出一种高效合并分块压缩感知方法(Multi-block Combined Compressed Sensing,MBCS),适用于基于焦平面阵列的压缩成像系统,它结合了并行采样和快速重建优势,可通过低分辨红外探测器实现低分辨并行测量和高分辨图像快速重建。与传统的基于压缩感知超分辨成像相比,该方法可提升高分辨图像重建的质量,同时实现高速重建。本文对光学系统原型和MBCS重建模型测量矩阵构建过程进行了研究,讨论了合并块大小对重建性能的影响,发现存在最优块大小使重建速度与重建质量都最优。此外,本文还实现了基于GPU加速的MBCS重建算法,用于进一步改进并行成像系统的图像重建速度。仿真和光学实验验证了该光学系统并行采样和快速重建策略的有效性,512×512分辨率成像与显示速度可达到5 Hz。Imaging sensors in medium and long-wave infrared spectrum are extremely expensive.Therefore,for most consumers,remote high-resolution imaging and real-time display in these spectrums are still a challenge.This paper proposes an effective block compressed sensing method called Multi-block Combined Compressed Sensing(MBCS)adapting to Focal Plane Array Compressed Imaging system(FPA CI),which combines parallel sampling and fast reconstruction.The high-resolution images can be reconstructed from low-resolution measurement results in real-time using a low-resolution infrared sensor.The results showed that,compared with the traditional CS-based super-resolution method,this method could greatly improve the quality of the reconstructed highresolution image and achieve a higher reconstruction speed.The optical prototype architecture and construction of the MBCS measurement matrix for the reconstruction model are also discussed.This study evaluated the reconstruction performance in terms of the block size and found that the optimal block size needed to consider both speed and reconstruction quality.Furthermore,the MBCS reconstruction algorithm with GPU acceleration was implemented to improve the image reconstruction speed of the highly parallel image system.In the experiment,the optical system and the strategy of rapid imaging and reconstruction were verified via simulation and optical experiments,which showed that the imaging speed of 512×512 resolution could reach 5 Hz.

关 键 词：压缩成像 分块压缩感知 中红外 焦平面阵列 图像处理单元 

分 类 号：TN215[电子电信—物理电子学]