面向傅里叶红外成像光谱仪干涉数据立方的场景非均匀性校正方法

In recent years, infrared detector array is replacing single-point detector and linear detector in Infrared Imaging Fourier-Transform Spectrometer based on the Michelson interferometer, which makes the spectrometer be capable of achieving spatial and spectral information simultaneously. However, infrared detector array also introduces non-uniformity noise to interferometric data cube, which reduces the measurement accuracy of the spectrum. Unfortunately, although existing correction methods for interferometric data cube is somehow effective, they all meet the same problem that the correction coefficient can not adaptively adjust according to the varying of non-uniformity, which causes their performance degradation along with time. To solve this problem, we plan to 1) research and analysis internal relations of the incident light spectrum, principle of interfere, time-varying non-uniformity and the interferometric signal characteristics, establish the interferometric data cube signal model based on time-varying non-uniformity; 2) transform the adaptive adjustment of correction coefficient into a optimization problem, and set the regularization term based on the characteristics of time-varying non-uniformity; 3) find a suitable iteration way for interferometric data cube according to its characteristics of interferometric signal and its data structure. Finally, a scene-based new nonuniformity correction method for interferometric data cube is proposed. Through this project, it is expected that not only will the theory of interferometric signal model be improve, but also the degradation problems of existing methods will be solved, which effectively promotes the accuracy and the reliability of Infrared Imaging Fourier-Transform Spectrometer in practical applications.

近年来，基于迈克仍逊干涉原理的傅里叶红外光谱仪开始采用面阵探测器取代传统的点、线探测器，可实时同步获取二维空间和一维光谱信息。但面阵探测器也会将非均匀性噪声引入干涉数据立方，造成光谱测量精度降低。现有干涉数据立方非均匀性校正方法虽然降低了非均匀性噪声，但均无法针对非均匀性的时变性进行自适应调整，存在校正效果随时间退化的问题。针对这一问题，课题拟从入射光光谱、干涉原理、时变非均匀性与干涉信号特性的内在联系入手，建立基于时变非均匀性的干涉数据立方信号模型；然后将校正系数的自适应调整抽象为最优化问题，构建时变非均匀性的正则化项；并探索与干涉信号特性、数据结构相适应的校正系数迭代方式，最终形成一种面向干涉数据立方的场景非均匀性校正新方法。通过课题的研究，不仅有望完善时变非均匀性影响下的干涉信号模型理论，还可解决现有方法校正效果退化的问题，有效提升傅里叶红外成像光谱仪在实际应用中的准确性和可靠性。

采用面阵探测器的傅里叶红外成像光谱仪可实时同步获取二维空间和一维光谱信息。但其中面阵探测器的非均匀性噪声会引入干涉数据立方，造成光谱测量精度降低。现有校正方法无法实现基于场景的非均匀性校正，存在校正效果随时间退化的问题。针对这一问题，课题从数据立方模型的构建、自适应场景非均匀性校正方法设计以及优化等方面展开研究，形成了一种面向干涉数据立方的场景非均匀性校正新方法。具体成果如下：1、构建了傅里叶红外成像光谱仪干涉数据立方信号模型，为校正方法的研究奠定了理论基础；2、提出了基于超分辨率与低秩表示的干涉数据立方预降噪方法，相比传统方法，能同时去除多种类型的噪声，降低各种噪声对非均匀性校正的干扰；3、提出基于稀疏表示与数据立方间迭代的非均匀性校正方法，采用稀疏表示对期望信号进行估计，根据数据立方的结构设计了迭代方式，实现了基于场景的非均匀性校正；4、通过可见光图像的融合与光谱的解混，区分了数据立方中的不同物质，优化了迭代的稳定性。项目成果在JOSAA、RS、AO、IEEE Access/TGRS等遥感或红外权威期刊发表SCI论文10篇，IEEE International Geoscience and Remote Sensing Symposium (IGARSS)上发表口头报告论文2篇；授权发明专利1项，申请发明专利1项；项目成果能有效降低红外探测器非均匀性响应效应对干涉信号的干扰，提高干涉数据质量，减小测量光谱的畸变，使得后续基于光谱指纹的物质分类算法能更加准确的对测量目标进行物质分类，为利用红外光谱技术在目标进行检测和识别等实际应用提供准确性和可靠性保障。