Partial Least Squares Processing of Near‐Infrared Spectra for Discrimination and Quantification of Adulterated Olive Oils

Abstract

A new processing based on partial least squares (PLS) algorithm for the discrimination and determination of adulterants in pure olive oil using near‐infrared (NIR) spectroscopy has been introduced. The 280 adulterations of olive oil with corn oil (n=70), hazelnut oil (n=70), soya oil (n=70), and sunflower oil (n=70) were prepared, and their NIR spectra in the region 12,000–4550 cm^?1 were collected. The 70 spectra of each adulteration of olive oil were divided into two sets, 50 spectra for a calibration set and 20 spectra for a prediction set. The spectra of a total calibration set (n=200) were separated into individual adulterant calibration sets (n_i=50, i=corn, hazelnut, soya, sunflower) by using discriminant PLS (DPLS) analysis, and PLS calibration models for the quantification of adulterants with corn oil, hazelnut oil, soya oil, or sunflower oil were developed separately. A variety of wavelength ranges and data pretreatments were examined for obtaining optimal results for the discrimination and quantification objects. Four PLS models for differentiating the adulterant types were evaluated by classifying the NIR spectra of a total prediction set (n=80) into known adulterant types. Then, these known adulterant spectra were analyzed by the PLS calibration models developed for each type to determine the content of an adulterant in pure olive oil. The results of evaluation revealed that the processing reported in this article works excellently for the discrimination and quantification of the adulterations of olive oil.     

近红外光谱技术结合递归偏最小二乘算法对土壤速效磷与速效钾含量测定研究

土壤速效磷与速效钾在近红外区没有直接与它们相关的吸收峰,只能借助与其他拥有直接吸收峰物质(有机质,碳酸盐,粘土矿物,水分等)之间的相关关系而被近红外光谱技术所预测。这种相关关系会随着土壤样品构成的不同而不断变化,因此采用固定结构的近红外光谱模型很难对速效磷与速效钾取得较好的预测效果。提出采用递归偏最小二乘法(RPLS)在预测过程中递归更新土壤速效磷与速效钾的回归系数,以提高模型的预测能力;比较了偏最小二乘法(PLS),局部加权PLS(LW-PLS),滑动窗口LW-PLS(LW-PLS2)和RPLS对于土壤速效磷与速效钾含量的预测结果。194份土壤样品根据土壤类型分为建模集与预测集：建模集包含120份人为土样品;预测集则包含29份铁铝土样品,23份人为土样品和22份初育土样品。结果表明：RPLS模型取得了最优的预测结果,获得的决定系数(R2)分别为0.61与0.76,预测相对分析误差 (RPD)分别为1.60与2.05。说明RPLS通过不断更新模型的回归系数,能够适应新加入建模集样品的信息。相比于其他方法,预测精度更高,适用范围更广。     

Quantitatively Determination of Available Phosphorus and Available Potassium in Soil by Near Infrared Spectroscopy Combining with Recursive Partial Least SquaresSCIEI北大核心CSCD

Soil available phosphorus (P) and available potassium (K) don't possess direct spectral response in the near infrared (NIR) region. They are predictable because of their correlation with spectrally active constituents (organic matter, carbonates, clays, water, etc.). Such correlation may of course differ between the soil sample sets. Therefore, the NIR calibration models with fixed structure are difficult to achieve good prediction performances for soil P and K. In this work, the method of recursive partial least squares (RPLS), which is able to update the model coefficients recursively during the prediction process, has been applied to improve the predictive abilities of calibration models. This work compared the performance of partial least squares regression (PLS), locally weighted PLS (LW-PLS), moving window LW-PLS (LW-PLS2) and RPLS for the measurement of soil P and K. The entire data set of 194 soil samples was split into calibration set and prediction set based on soil types. The calibration set was composed of 120 Anthrosols samples, while the prediction set included 29 Ferralsols samples, 23 Anthrosols samples and 22 Primarosols samples. The best prediction results were obtained by the RPLS model. The coefficient of determination (122) and residual prediction deviation (RPD) were respectively 0.61, 0.76 and 1.60, 2.05 for soil P and K. The results indicate that RPLS is able to learn the information from the latest modeling sample by recursively updating the model coefficients. The proposed method RPLS has the advantages of wider applicability and better performance for MR prediction of soil P and K compared with other methods in this work.     

基于近红外光谱和最小二乘支持向量机的聚丙烯酰胺类型鉴别

提出了一种基于近红外光谱分析技术和最小二乘支持向量机的鉴别方法,能够快速、无损鉴别聚丙烯酰胺的三种类型。获取非离子,阴离子和阳离子等三种类型的聚丙烯酰胺样本的近红外漫反射光谱,用主成分分析方法对样本光谱数据进行降维,并提取主成分。基于前三个主成分对三种类型的聚丙烯酰胺样本进行聚类分析,并将主成分作为最小二乘支持向量机的输入。通过基于网格搜索的交叉验证方式优化最小二乘支持向量机的参数和作为其输入的主成分个数。每种类型聚丙烯酰胺各采集60个样本,共采集180个样本,每种类型样本随机选取45个样本,共135样本作为训练样本集,剩余45个样本作为测试集。为了验证该方法能否鉴别掺假样本,制备了掺入不同比例非离子聚丙烯酰胺的5个阴离子和5个阳离子聚丙烯酰胺样本。采用基于训练样本集交叉验证预测误差的F统计显著性检验方法来确定样本的鉴别结果误差阈值。结果表明,预测测试集时,准确率为100%。预测10个混和样本时,所有混合样本都被准确识别出。说明该方法能快速无损鉴别不同类型的聚丙烯酰胺并且具有掺假鉴别能力,为聚丙烯酰胺类型的快速鉴别提供了一种新方法。     

Noninvasive hemoglobin measurement based on optimizing Dynamic Spectrum method

An optimizing Dynamic Spectrum method based on short-time Fourier transform for noninvasive hemoglobin measurement was proposed. A total of 187 volunteers’ spectral data and blood samples were prepared for the calibration (167 samples) and prediction (20 samples) sets. Data processing with optimizing dynamic spectrum method was set as experimental group, while control group was with classical method based on fast Fourier transform. The model for hemoglobin prediction was established with partial least square regression. In the experimental group, the average prediction correlation coefficient was 0.8399, while that of the control group was 0.6686. The results demonstrate that the optimizing method has a good accuracy improvement for noninvasive hemoglobin measurement. The optimizing dynamic spectrum method combining statistical characteristics analysis as a modified analytical method has a promising application prospect of application in noninvasive blood component measurement field.     

Vis-NIR光谱模式识别结合SG平滑用于转基因甘蔗育种筛查

以Savitzky-Golay (SG)平滑筛选,主成分分析(PCA)分别结合有监督的线性判别分析(LDA)、无监督的系统聚类分析(HCA),应用于转基因甘蔗育种筛查的可见-近红外(Vis-NIR)无损检测。提出兼顾随机性、稳定性的定标、预测、检验框架;取田间种植处于伸长期甘蔗叶样品456个,具有Bt基因和Bar基因的转基因样品(阳)306个,非转基因样品(阴)150个;随机选取156个为检验集(阴性50、阳性106),余下为建模集(阴性100、阳性200,共300),建模集再随机划分为定标集(阴性50、阳性100,共150)、预测集(阴性50、阳性100,共150)共50次;扩充SG平滑点数,同时删除绝对值偏小的高阶导数模式,共264个平滑模式用于模型筛选;采用前3个主成分两两组合,再根据模型效果选出最优主成分组合;基于所有定标、预测集划分和SG平滑模式,建立SG-PCA-LDA和SG-PCA-HCA模型,根据平均预测效果优选参数,使模型具有稳定性;最后用检验集进行模型检验。经SG平滑后,PCA-LDA和PCA-HCA的建模精度、稳定性均显著改善;最优SG-PCA-LDA模型阳性、阴性样品检验识别率分别达到94.3%和96.0%;最优SG-PCA-HCA模型阳性、阴性样品检验识别率分别达到92.5%和98.0%。结果表明：Vis-NIR光谱模式识别结合SG平滑可用于转基因甘蔗叶的准确识别,提供了一种简便的转基因甘蔗育种筛查方法。     

可见近红外高光谱成像对灵武长枣定量损伤等级判别

利用可见近红外（Vis-NIR）高光谱成像技术对完好和损伤等级灵武长枣进行快速识别检测。采用定量损伤装置得到损伤Ⅰ,Ⅱ,Ⅲ,Ⅳ和Ⅴ级的灵武长枣,借助高光谱成像系统采集完好长枣和损伤长枣样本高光谱图像。提取感兴趣区域（region of interest,ROI）并计算样本平均光谱值。利用光谱-理化值共生距离算法（SPXY）将420个长枣样本按3∶1的比例划分校正集315个和预测集105个。灵武长枣原始光谱建立偏最小二乘判别分析（PLS-DA）分类模型,得到校正集和预测集准确率分别为72.70%和86.67%;灵武长枣原始光谱数据采用移动平均（MA）、卷积平滑（SG）、多元散射校正（MSC）、正交信号修正（OSC）、基线校准（baseline）和去趋势（de-trending）等方法进行光谱预处理并建立PLS-DA分类判别模型。通过分析比较,得到MSC-PLS-DA为最优分类判别模型,校正集准确率为76.19%,预测集准确率为86.67%,其中校正集比原始光谱建模准确率提高了3.49%,预测集准确率较原始光谱建模结果未提高;为了提高建模效果,对灵武长枣原始光谱和预处理后的光谱分别采用连续投影算法（SPA）、无信息变量消除（UVE）、竞争性自适应加权抽样（CARS）和区间变量迭代空间收缩法（iVISSA）等算法提取特征波长,建立PLS-DA分类判别模型,结果表明,MSC-CARS-PLS-DA为最优模型组合,校正集准确率为77.14%,预测集准确率为89.52%,建模准确率较原始光谱建模准确率分别提高了4.44%和2.85%。结果表明,Vis-NIR高光谱成像技术结合MSC-CARS-PLS-DA模型可实现灵武长枣损伤等级的快速识别。     

基础数据准确性对近红外光谱分析结果的影响

基础数据的准确性是影响近红外光谱分析结果的一个重要因素。文章以人工配制的四组分混合物体系和实际的汽油校正集样本为例,通过人为增加基础数据误差的方法,研究了基础数据的准确性对近红外光谱分析结果的影响。结果表明,基础数据的准确性对近红外分析模型及其预测结果都有一定的影响,基础数据越准确,所建立模型的精度越高,其对未知样本的预测结果也越准确。对于精度相对较差测试方法提供的基础数据,通过大量样本的光谱分析和化学计量学统计处理,近红外方法有可能得到更精确的预测结果。     

Multivariate calibration in TXRF analysis of water

The aim of this paper is the application of multivariate linear calibration for quantitative determination of elements (K, Cd, Co, Hg, As, Pb, Ni, and Al) in water by using Total Reflection X‐ray Fluorescence Analysis with partial least squares (PLS) as a regression method to improve a result of common univariate method. In purpose of elimination of matrix effects in X‐ray fluorescence analysis, experimental design was applied. As a set of standard samples for multivariate calibration, a five‐level eight‐factor calibration design of 25 samples was chosen, ensuring mutual orthogonality of factors. For model's validation, the independent test set of 15 samples was examined. The collection of spectra and quantitative measurements was carried out on S2 PICOFOX. The PLS regression was performed by using software package STATISTICA. Quality indicators of multivariate calibration as slope (b) and intercept (a) of calibration, correlation coefficient (r), determination coefficient (R²), root mean square errors of calibration and of prediction, standard errors of calibration and of prediction, biases of calibration, and biases of prediction were calculated. These results were compared with the univariate model, and as a result, the multivariate calibration method exceeds the univariate one. The obtained results could be applied in a laboratory for an analysis of water solutions in the concentration range 0.05–2.00 mg/L. In many real situations, when analytical chemist deals with multi‐element mixtures, multivariate calibration approach combined with orthogonal design for multivariate calibration set could be successfully used to improve a conventional univariate calibration.     

A measurement of the 50Ti(γ, n) and 50Ti(γ, n0) cross sections

Measurements of the ⁵⁰Ti(γ, n) and ⁵⁰Ti(γ, n₀) cross sections have been made in the energy range of the giant dipole resonance (GDR). Assuming the GDR is split into two isospin components, approximated as Lorentzians, a calculation based on statistical decay of the GDR states is consistent with the experimental results.     

连续偏振光谱的稻种发芽率检测方法研究

针对目前传统稻种发芽率检测方法周期长、精度低的问题,提出新颖的基于连续偏振光谱技术实现稻种发芽率快速、无损检测的方法。以不同老化天数稻种为检测目标,10 min为检测时间点,使用起偏器将光纤准直光源调制成线偏振光垂直入射稻种浸出液,而后以5°为间隔旋转检偏器,并通过光纤光谱仪检测透射的光谱,对检测的偏振光谱通过归一化预处理后,根据不同发芽率稻种检测时偏振角及波长的贡献给出特征偏振角和特征波长,特征偏振角为0°,5°和25°,特征波长为576,620和788 nm,将获取的连续偏振光谱以特征偏振角和特征波长处的透射率为输入,构建稻种发芽率检测模型。分别比较运用偏最小二乘法回归(partial least squares regression,PLSR)、BP神经网络(back propagation neural network,BPNN)、径向基神经网络(radial basis function neural network,RBFNN)三种建模方法建立稻种发芽率检测模型。分别用老化天数为0,2,4,6 d的稻种,在不同的偏振角共测量1 520组实验数据,其中912组数据作为校正集,608组数据作为预测集,建模结果表明三种模型预测精度较高,其中RBFNN模型预测精度最高,其相关系数r为0.976,均方误差RMSE为0.785,平均相对误差MRE为0.85%。表明利用连续偏振光谱技术通过多维度光谱信息能够有效实现稻种发芽率的快速、准确检测。     

NIR光谱结合稳定等效MW-PLS方法用于高脂血症指标分析

基于稳定性、等效性对移动窗口偏最小二乘(MW-PLS)方法进行改进,应用于高脂血症指标总胆固醇(TC)、甘油三酯(TG)无试剂近红外光谱分析的波长优化。提出兼顾随机性、稳定性的定标、预测、检验框架。从全体人血清样品(阴性145、阳性158,共303)中,随机选取103个为检验集(阴性44、阳性59),余下为建模集(阴性101、阳性99,共200),再将建模集随机划分为定标集(阴性51、阳性49,共100)、预测集(阴性50、阳性50,共100)共50次;基于所有划分的平均预测效果优选模型参数,使得模型具有稳定性;采用样品优选的模型进行检验。TC,TG最优MW-PLS波段分别为1 556～1 852,1 542～1 866 nm;为了解决由于材料性能、成本的因素对仪器分光系统设计的制约,提出等效模型集,并得到TC和TG等效模型集的唯一公共波段1 542～1 852 nm。检验结果表明：采用最优MW-PLS波段,检验样品的TC和TG预测均方根误差(V_SEP)分别为0.177和0.100 mmol L-1、相关系数(V_R_P)分别为0.988和0.996,高脂血症的灵敏度、特异性分别为95.0%和90.5%;基于公共等效波段,TC和TG的V_SEP值分别为0.177和0.101 mmol·L-1、V_R_P值分别为0.988和0.996,灵敏度、特异性分别为92.7%和90.3%。结论：近红外光谱结合稳定等效MW-PLS方法提供了一种有潜力的大人群血脂检测工具。      

New functional evaluation methods of sound insulation system using the SEA method and an Lx evaluation criterion

Recently, for the purpose of reducing residential environmental noise, many sound insulation systems have often been improved acoustically by changing their geometrical scales and/or acoustical characteristics. In this paper, new functional evaluation and probabilistic prediction methods for these improvements are theoretically and experimentally proposed in practical expression forms by introducing a few functional parameters. These functional parameters introduced only for the prediction are supported by many physical structural factors closely related to the well-known statistical energy analysis method, and are easily estimated in a preliminary experiment. The estimation procedures developed are based on two error criteria using the actual overall frequency band data. The used least-squares error criterion is the most fundamental method and the L_x evaluation criterion matches the actual situation of estimating the representative evaluation indices. Finally, by using musical sound as an input noise, the effectiveness of the proposed method is experimentally confirmed by applying it to some actual problems.