基于混叠编码感知的高分辨光谱视频获取机理与方法研究

High-resolution spatial-temporal-spectral imaging is a key issue to the fields of military reconnaissance, seismic monitoring and medical diagnostic. Traditional spectral imaging method can not obtain high-resolution spatial-temporal-spectral multi-dimensional information simultaneously because of imaging principle defect. This project proposes a new spectral video acquisition method based on aliasing coding perception with compressive sensing and computational spectral imaging. The method makes high-resolution spatial-temporal- spectral information no longer restrict to the density of the spectral detector. The main research content of this project includes how to obtain spatial-temporal-spectral information collaboratively, excavating and representing multi-dimensional prior information, and inversion decoding with the prior information. The main innovations of this project are listed below: 1. in the encoding perceptual stage, and we combine coded aperture, coded exposure and spectral aliasing approaches and take advantages of jointing them. 2. in the inversion decoding stage, integrate local and nonlocal sparsity into a unified variational framework, which guarantees the reconstruction accuracy of spectral video. This project is a breakthrough of current computational imaging and compressive sensing theory, and a milestone of high resolution spectral detection. Its technical achievements will promote the applications of the high-spectral imaging to many fields, such as high-speed object recognization, sensitive areas surveillance and medical diagnostic etc, and improve the performance significantly.

高分辨空时谱多维信息光谱成像在军事侦察、地震监测、医疗诊断等领域有迫切需求。传统光谱数据获取方法从原理上难以兼得空间、时间、谱间的高分辨率。本课题将将压缩感知与计算光谱成像技术相结合，提出基于混叠编码感知的高帧率光谱数据获取方法。该方法将使得光谱图像的空间和光谱分辨率不再受制于探测器的点阵密度，是一种全新的光谱成像模式。课题主要研究如何实现空-时-谱信息的高分辨协同获取，探索多维信号先验信息的挖掘和表达方法，探究利用先验信息反演解码的策略。课题创新点：1.编码感知阶段，将孔径编码、曝光编码、光谱混叠等技术有机结合，能够在观测数据中尽可能多地保留高分辨信息；2.反演解码阶段，把局部和非局部稀疏约束整合到统一的变分框架中，保证了光谱视频的重建精度。本课题预期在理论上有突破，技术上有创新，为推动探测成像在高速机动目标识别、敏感区域侦察、医疗诊断等传统方法难以胜任的场合的应用奠定理论和技术基础。

本课题面向高分辨多维信息光谱成像在军事侦察、地震监测、地质勘探等领域的迫切需求，针对传统光谱数据获取方法从原理上难以兼得空间、时间、谱间的高分辨率的难题，基于压缩感知和计算成像技术，提出了基于混叠编码感知的高帧率光谱数据获取方法。该方法使得光谱图像的空间和光谱分辨率不再受制于探测器的点阵密度，是一种全新的光谱成像模式。课题主要研究了如何实现空-时-谱信息的高分辨协同获取，探索了多维信号先验信息的挖掘和表达方法，以及利用先验信息反演解码的策略。本课题创新点：1.编码感知阶段，将孔径编码、曝光编码、光谱混叠等技术有机结合，能够在观测数据中尽可能多地保留高分辨信息；2.反演解码阶段，把空间稀疏性和谱间相关性整合到统一的优化框架中，保证了光谱数据的重建精度。本课题在理论上有突破，技术上有创新，为推动探测成像在识别高速机动目标、监视侦察敏感区、掌握实时多变态势等传统方法无法胜任的场合的应用奠定理论和技术基础。