基于经验模态分解的干涉图滤波方法  

Interferogram Filtering Method Based on Empirical Mode Decomposition

作　　者：景娟娟[1,2] 相里斌[2] 李然[1] 石大莲[1] 

机构地区：[1]中国科学院西安光学精密机械研究所,中国科学院光谱成像技术重点实验室,陕西西安710119 [2]中国科学院光电研究院,中国科学院计算光学成像技术重点实验室,北京100094

出　　处：《光学学报》2013年第10期62-67,共6页Acta Optica Sinica

基　　金：国家973计划(2009CB724005)

摘　　要：干涉图滤波是干涉光谱成像仪光谱反演过程中的一个关键环节,常用的滤波方法主要是差分法和拟合法。差分法对背景噪声不能完全去除;拟合法则需要先验知识,而且在干涉数据两端拟合误差较大。经验模态分解(EMD)方法是近年来提出的一种新的用于线性和稳态谱分析信号处理方法,该方法提出后在很多领域得到广泛应用。将EMD方法应用到干涉图的滤波过程中,使得对背景噪声的提取更为合理,而且具有自适应性,避免了常用滤波方法的不足。利用实验室实际获取的数据进行分析,可以看出:EMD滤波后空间维的光谱相对均方根误差(RQE)均值为0.0068,精度最高;其次为拟合法,RQE均值为0.0073;最后为差分法,RQE均值为0.0079。Interferogram filtering is a key technique in the process of spectral recovery of imaging Fourier transform spectrometer. Differential filtering and polynomial filtering are usually used, but differential filtering cannot filter the noise completely, and polynomial filtering, which needs the noise type when filtering, produces big bias at both ends of the interferograms. Empirical mode decomposition （EMD）, a new signal processing method in linearity and stationary spectral analysis, is now widely used in many aspects. EMD technique is applied to interferogram filtering. It can avoid the disadvantages of differential filtering and polynomial filtering, and it is more reasonable to extract the background noise. The data acquired in the laboratory are used to analyze the precision of different filtering methods. The result indicates that the precisions of differential filtering, polynomial filtering and EMD are 0. 0079, 0. 0073, 0. 0068, respectively. EMD is the optimum filtering method, followed by the polynomial filtering and differential filtering.

关 键 词：光谱学 光谱成像 干涉图 滤波 经验模态分解 光谱反演 

分 类 号：TH744.1[机械工程—光学工程]