光谱基线校正算法研究与应用进展  

Research and Application Progress of Algorithms for Spectral Baseline Correction

作　　者：王海朋 褚小立[1] 陈瀑[1] 刘丹[1] 李敬岩[1] 许育鹏[1] WANG Hai-Peng;CHU Xiao-Li;CHEN Pu;LIU Dan;LI Jing-Yan;XU Yu-Peng(Research Institute of Petroleum Processing,SINOPEC,Beijing 100083,China)

机构地区：[1]中石化石油化工科学研究院,北京100083

出　　处：《分析化学》2021年第8期1270-1281,共12页Chinese Journal of Analytical Chemistry

基　　金：国家重点研发计划项目(No.2017YFB0306501)资助。

摘　　要：基线漂移是拉曼、中红外、近红外以及激光诱导击穿光谱等光谱仪器测量过程中经常出现的问题,会对光谱的定量和定性分析产生不利影响。基线校正是利用数学算法对光谱数据进行预处理的过程,以便于后续的尤其是结合化学计量学方法的光谱定性和定量分析。依据算法原理,常见的基线校正算法包括导数法、迭代多项式拟合法、分段拟合的算法、移动窗口平滑算法、小波变换算法、基于惩罚最小二乘的算法、鲁棒基线估计法。这些算法可在很大程度上消除基线漂移给定量和定性分析带来的不利影响,但各有特点和不足。近年来,针对上述基线校正算法的不足,研究者相继开发了一些改进算法和新型的光谱基线校正算法。改进的算法有自适应最大和最小多项式拟合荧光背景扣除算法、基于子空间向量夹角判定的区间线性拟合算法、动态移动最小二乘多项式平滑算法、确定最优小波分解尺度的能量分布法、自适应迭代重加权惩罚最小二乘算法等;新型的算法包括基于荧光褪色效应原理的荧光褪色差分法、基于图像处理中形态学算子的基线校正方法、基于稀疏表示的同时拟合纯光谱和基线的方法等。这些改进的和新型的基线校正算法进一步改善了光谱质量,提高了光谱定量和定性分析的准确性和稳健性。本文系统评述了光谱基线校正的基本算法、改进算法和新型算法及其应用研究进展,并对其发展趋势进行了展望。The baseline drift is common in the measurement process of Raman spectroscopy,middle infrared(MIR)spectroscopy,near infrared(NIR)spectroscopy,laser-induced breakdown spectroscopy(LIBS)and other spectral instrumentation,which will seriously deteriorate the result of subsequent qualitative and quantitative analysis of spectra.To obtain accurate and clear results,some effective baseline correction algorithms have been implemented to correct the baseline of spectra before subsequent qualitative and quantitative analysis,especially combined with chemometric methods.Baseline correction algorithms are mainly derivative method,iterative polynomial fitting,piecewise fitting,moving window smoothing,wavelet transform(WT),penalized least squares and robust baseline estimation(RBE).These algorithms can eliminate the adverse effects of baseline drift on quantitative and qualitative analysis to a large extent,but each of them has some shortcomings in certain aspects.In recent years,in response to the drawbacks of the aforementioned algorithms,some improved and novel algorithms for baseline correction have been proposed one after another.Improved algorithms are mainly adaptive minmax polynomial fitting fluorescence background subtraction algorithms,interval linear fitting based on subspace vector angle,dynamic moving Savitzky-Golay algorithms,selection method of optimum decomposition in wavelet transform based on energy distribution,adaptive iteratively reweighted penalized least squares(airPLS),etc.Novel algorithms include fluorescence photo-bleaching difference approach(FBDA),algorithms based on morphological operators,synchronous fitting algorithms of pure spectrum and baseline based on sparse representation,etc.These algorithms have not only improved the quality of spectra but also further enhanced the accuracy and robustness of subsequent quantitative and qualitative analysis based on spectra.This article systematically reviews basic algorithms,improved algorithms and novel algorithms for spectral baseline correction,as wel

关 键 词：化学计量学 基线校正 拉曼光谱 激光诱导击穿光谱 迭代多项式拟合 分段拟合 惩罚最小二乘 评述 

分 类 号：O657.3[理学—分析化学]