高分辨光学压缩光谱成像方法与实验研究  

Research on Principle and Experimentation of High-Resolution Optical Compressive Spectral Imaging

作　　者：陈宇恒[1,2,3,4] 周建康[1,2,3,4] 陈新华[1,2,3,4] 季轶群[1,2,3,4] 沈为民[1,2,3,4] 

机构地区：[1]苏州大学现代光学技术研究所,江苏苏州215006 [2]江苏省先进光学制造技术重点实验室,江苏苏州215006 [3]江苏省现代光学技术重点实验室,江苏苏州215006 [4]教育部现代光学技术重点实验室,江苏苏州215006

出　　处：《光学学报》2014年第1期107-112,共6页Acta Optica Sinica

基　　金：国家自然科学基金(61205188);高等学校博士学科点专项科研基金(20113201110016)

摘　　要：光学压缩光谱成像是融合了压缩感知原理的新型光谱成像技术,具有降低数据采集量、能对景物实施凝视拍摄、提高信噪比等优点。考虑到采样质量对最终成像质量的影响,在现有成像系统中均采用采样间隔与调制间隔匹配的方法,但此方法降低了系统的采样率,损耗了原始光谱分辨率。针对上述成像方法缺陷,克服采样间隔和调制间隔匹配的成像系统设计要求,所设计实验装置使其光谱分辨率理论值提高至原先方法的3倍以上,并对最优化方法进行改进,在正则化函数中增加表征数据光谱维连续性的变差项,增强数据重建可控性及可靠性。实验结果表明,新方法下实验装置的光谱维通道数提升,各波段图像和特定位置光谱曲线能精确反映目标物的空间特性和光谱特性。Optical compressive spectral imaging method is a novel spectral imaging technique that draws in the inspiration of compressed sensing, which has the features such as reducing acquisition data amount, realizing snapshot imaging for certain scenery, increasing signal to noise ratio and so on. Considering the influence of the sampling quality on the ultimate imaging quality, matching the sampling interval with the modulation interval in the former reported imaging system, while the depressed sampling rate leads to the loss on the original spectral resolution. To overcome that technical defect, the demand for the matching between sampling interval and modulation interval is disposed and the spectral resolution of the designed experimental device increases more than threefold comparing with that of the previous method. Optimization method is improved and a variation term that represents the spectral-dimension continuousness of the data is added to the regularization function, which enhances the controllability and reliability for the data reconstruction. Result proves that the spectral channel number increasesto a great extent effectively, the average spectral resolution reaches 1 nm, and the spectral images and curves are able to perform the spatial and spectral character of the target accurately.

关 键 词：光谱成像 压缩感知 最优化方法 采样率 光谱分辨率 

分 类 号：O433.4[机械工程—光学工程]