干涉光谱技术在大气遥感、 天文探测及地物勘察等诸多领域的应用是当前国内外研究热点, 光谱重构作为遥感数据处理的重要环节与探测精度紧密相关。 干涉数据由于有限光程差采样导致复原光谱出现频率泄露, 切趾函数在光谱重构过程中可以起到平滑光谱、 保持复原光谱和其他类型分光技术探测光谱一致性的作用, 但同时会造成重构光谱的分辨率下降。 已有研究表明切趾函数并没有提高反演精度, 同时多个典型大气遥感载荷地面数据处理过程中并未使用切趾函数。 空间外差光谱技术由于其诸多优点在国内外引起广泛关注, 中科院安光所基于该技术成功研究出用于大气CO2探测的原理样机。 信噪比是光谱仪的核心指标之一, 从信噪比、 光谱分辨率和探测精度之间的关系出发研究切趾函数在干涉数据光谱重构中的影响。 针对当前传统切趾函数并没有达到最优旁瓣抑制效果, 以诺顿-比尔切趾函数为基础, 在分辨率降低相同的情况下, 获取最大的旁瓣抑制程度为判据构造了10种不同光谱展宽程度的切趾函数。 利用SCIATRAN辐射传输模型分析了大气CO2遥感探测中不同气体浓度造成的大气层顶的辐亮度差异, 推导了典型条件下不同光谱分辨率满足1%探测精度需求的信噪比要求。 以实验室空间外差光谱仪样机参数为基础, 通过仿真干涉数据和本文构造切趾函数分析了不同切趾程度下光谱分辨率和信噪比的变化关系。 最后利用实验室研制的样机开展了实验验证, 通过观测稳定均匀积分球辐射源获取干涉数据在不切趾的情况下计算信噪比, 以及干涉数据进行不同程度切趾后复原光谱信噪比。 仿真和试验结果表明干涉数据由于切趾对噪声的抑制信噪比逐渐升高, 同时造成光谱分辨率逐渐下降, 而探测精度由于分辨率下降对光谱信噪比的要求也逐渐升高。 探测精度对信噪比的需求提高明显高于切趾作用下光谱信噪比的升高趋势, 仿真数据和实测数据信噪比分别在切趾程度大于1.6倍和1.8倍的情况下低于探测精度对仪器信噪比需求, 即白噪声在噪声中占主导的情况下不切趾更有利于探测精度的保障。 该研究结果可以作为干涉数据光谱重构的参考。

The application of interferometric spectroscopy in many fields, such as atmospheric remote sensing, astronomical exploration and geophysical prospecting, is a research hotspot at home and abroad. Spectral reconstruction, as an important part of remote sensing data processing, is closely related to the detection accuracy. Interferogram due to limited optical path difference sampling results in frequency leakage in the restored spectrum. The apodization function can play a role in smoothing the spectrum, maintaining the spectral consistency of the interferometric spectroscopy and other spectroscopy techniques, but at the same time, the resolution (Full Width at Half Maximum, FWHM) of the reconstructed spectrum will be reduced. It has been shown that the apodization function does not improve the inversion accuracy, and the apodization function is not used in the ground data processing of several typical atmospheric remote sensing loads. Spatial heterodyne spectroscopy (SHS) has attracted wide attention at home and abroad due to its many advantages. Based on this technology, Anhui Institute of Optics and Fine Mechanics of Chinese Academy of Sciences has successfully developed a prototype for atmospheric CO2 detection. Signal-to-noise ratio (SNR) is one of the key indicators of spectrometer. This paper studies the influence of apodization function on spectral reconstruction of interferogram from the relationship between SNR, spectral resolution and detection accuracy. In view of the fact that the traditional apodization function does not achieve the optimal side-lobe suppression, this paper constructs ten apodization functions with different spectral extending based on Norton-Beer apodization function and the criterion of obtaining the maximum side-lobe suppression under the same resolution reduction. Radiative transfer Model of SCIATRAN was used to analyze the difference of atmospheric top radiance caused by different gas concentration in atmospheric CO2 remote sensing. The spectral SNR requirement of different spectral resolution meets the requirement of 1% detection accuracy under typical conditions was calculated. Based on the parameters of the laboratory prototype, the relationship between spectral resolution and SNR under different extending was analyzed by simulating interferogram and constructed apodization function in this paper. Finally, the experimental verification was carried out by using a prototype developed by the laboratory. The interferograms were obtained by observing the stable uniform integrating sphere, and the SNR was calculated without apodization, and the SNR was calculated after different apodization extending. The simulation and experimental results show that of the SNR is gradually increased due to the reduction of noise by apodization, and the spectral resolution is gradually reduced, while the requirement of SNR for detection accuracy is gradually increased due to the decrease of resolution. The spectral SNR requirement of detection accuracy is obviously higher than that under the apodization. The SNR of simulation data and measured data is lower than that of the requirement of detection accuracy when apodization extendings are greater than 1.6 times and 1.8 times, respectively. The need for the SNR of the instrument, that is, the white noise is dominant, is not conducive to the detection accuracy. The results of this paper can be used as a reference for spectral reconstruction, and the influence of inversion accuracy will be further analyzed in the future.