Laser & Optoelectronics Progress, Volume. 62, Issue 1, 0100011(2025)
Application of Infrared Spectroscopy Combined with Chemometrics in Material Evidence Analysis
Figures & Tables(11)
Fig. 2. Discriminant function diagrams of PCA+FDA model. (a) Raw spectral data; (b) first derivative spectral data; (c) fusion spectral data
Fig. 3. Principal component analysis of camellia oil and adulterated oil. (a) MIR spectra; (b) NIR spectra
Fig. 4. PLS regression of predicted vs actual soybean oil content in pure camellia oil samples. (a) ATR-MIR spectra; (b) FODR-NIR spectra
Fig. 5. Comparative ATR-FTIR spectra of asian elephant, indian leopard, and royal Bengal tiger
Fig. 6. PLS-DA plot to discriminate the blood spectra of asian elephant, indian leopard, and royal Bengal tiger
Table 1. Several spectral preprocessing methods commonly used in infrared spectroscopy
View tableTable 1. Several spectral preprocessing methods commonly used in infrared spectroscopy
Aim Method Explanation Data enhancement transform Mean centering Before establishing a model， it is often necessary to adopt a data enhancement algorithm to eliminate redundant information in the collected data， so as to enhance the differences between data samples， so as to effectively improve the representativeness and prediction ability of the built model Autoscaling Normalization Smoothing Savitzky-Golay convolution smoothing Smoothing is to eliminate the random noise in the spectral signal and improve the signal to noise ratio of the sample signal， which can improve the prediction effect of the model to a certain extent Moving window smoothing Baseline correction 1st derivative In spectrum preprocessing， derivative algorithm is mainly used to analyze the rate of change of spectral data， which is very effective for eliminating baseline drift， smoothing background interference and resolving overlapping peaks 2nd derivative Scattering correction Standard normal variate （SNV） They are mainly used to eliminate the influence of scattering on the spectrum caused by uneven particle distribution and particle size difference Multiplicative scatter correction（MSC）
Table 2. Comparison of chemical pattern recognition methods commonly used in infrared spectroscopy
View tableTable 2. Comparison of chemical pattern recognition methods commonly used in infrared spectroscopy
Classification Pattern recognition method Basic idea Characteristic Range of application Limitation Reference Supervised PLS-DA By way of projection， the predictor and observed variables are projected into a new space， and then a linear regression model is found in this new space to achieve the classification or discrimination task Noise filtering， dimensionality reduction， discrimination and classification， sample prediction， good interpretation ability The situation where the number of explanatory variables is large and there is multicollinearity， the number of sample observations is small and the interference noise is large When there is a high correlation between variables， PLS-DA may overfit the data ［ 11 -12 ］LDA Maximize the distance between classes and minimize the distance within classes Dimensionality reduction， discrimination and classification， high computational efficiency Sample data with clear class label， high dimensional feature space and approximate Gaussian distribution It is not suitable for samples with non-Gaussian distribution and cannot handle nonlinear problems ［ 13 -15 ］Supervised KNN By majority rule， a sample belongs to a class if most of its neighbors belong to that class Noise filtering， discrimination and classification， sample prediction， no training required Multiple classification problems， nonlinear problems and rare event classification Large computational complexity and sensitivity to distance measurement ［ 16 ］SVM An optimal hyperplane is found to divide the samples so that the distance between the two types of samples closest to the plane is maximized Discrimination and classification， sample prediction Binary classification problem， small sample high-dimensional data set， separable data set Sensitive to parameter adjustment and kernel function selection ［ 17 -18 ］SIMCA PCA analysis is used to build a model for each class， calculate the distance between the unknown samples and these models， and determine the category of the unknown samples according to the distance discrimination Dimensionality reduction， discrimination and classification， sample prediction Samples with clear category labels Large computational complexity ［ 19 -20 ］Non-supervised PCA Using a small number of basic principal components （PCS） to explain the correlation between a large number of variables， the dimensionality of the original data set is reduced and the original information content is preserved as much as possible Dimensionality reduction， sample classification， maximum retention of original variable information Data sets with a large number of features and sufficient sample size， especially when dimensionality reduction is needed to simplify the problem When the sample size is insufficient or there are too many variables， PCA may not be effective in extracting representative principal components ［ 21 -22 ］HCA The variables or samples that are less similar are aggregated into one class， and the variables or samples that are more similar are aggregated into one class Discrimination and classification， intuitiveness and visualization Samples that are not clearly labeled or classified It is sensitive to distance calculation methods and feature weights and difficult to deal with large data sets ［ 23 ］
Table 3. Summary of chemometric methods used in physical evidence analysis
View tableTable 3. Summary of chemometric methods used in physical evidence analysis
Research object Sample Aim Analytical technique Applied statistical method Reference Trace evidence Household paints Classification FTIR PCA ［ 27 ］Textile fibres Identification ATR-FTIR PCA ［ 28 ］Automobile lampshades Classification FTIR PCA-FDA ［ 29 ］Explosives Identification NIR PLS-DA ［ 30 ］Explosives Identification NIR PLSR ［ 31 ］Physical evidence of food， medicine， and environmental factors Illicit liquors Origin identification ATR-FTIR LDA ［ 32 ］Crocus sativus Origin identification NIR、MIR PLS-DA ［ 33 ］Sika deer antler cap powder Adulteration detection FTIR PCA-SVM ［ 34 ］Camellia oils Adulteration detection MIR-ATR，FODR-NIR SIMCA，PLS ［ 35 ］Soil Pollutant concentration NIR PLS ［ 36 ］Biological evidence Blood Species differentiation ATR-FTIR PLS-DA ［ 38 ］Blood Gender and racial discrimination ATR-FTIR PLS-DA ［ 39 ］Blood Prediction of the formation time of bloodstains ATR-FTIR MLR，PLSR ［ 40 ］Hair Identification ATR-FTIR PLS-DA ［ 41 ］Seminal fluid Classification and identification ATR-FTIR PCA-LDA，PLSR ［ 42 ］Document evidence Erasable pens Classification FTIR PCA ［ 44 ］Black toner Classification μ-FTIR PCA-HCA ［ 45 ］Banknotes Identification NIR SIMCA，SPA-LDA ［ 46 ］Papers Identification ATR-FTIR PCA ［ 47 ］Paper express document bags Classification FTIR PCA，FDA ［ 48 ］
Tools
Get Citation
Copy Citation Text
Ruiting Feng, Hongda Li, Zhichao Yang, Yumu Liu. Application of Infrared Spectroscopy Combined with Chemometrics in Material Evidence Analysis[J]. Laser & Optoelectronics Progress, 2025, 62(1): 0100011
Paper Information
Category: Reviews
Received: Apr. 15, 2024
Accepted: May. 22, 2024
Published Online: Jan. 3, 2025
The Author Email:
CSTR:32186.14.LOP241107
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20