Research On Data Processing Methods Of Surface Enhanced Raman Spectroscopy

Raman spectroscopy is a kind of molecule scattering spectroscopy.Analytical advantages of rapid response time,non-contact measurement,less detection restrictions and high selectivity make it a powerful technique for analyzing various substances,such as organic substance,inorganic substance and complicated mixtures.Since Raman scattering surface of the material molecules is small,the intensity of Raman spectra is very weak,it is more difficult for detection and analysis of Raman spectroscopy.With the specially processed substrate based on a rough metallic surfaces or metallic sols,the Surface Enhanced Raman Scattering?SERS?can enhance the intensity of the spectrum representing the molecular information by 10⁶¹0¹⁵ times,and thus obtains extremely high sensitivity.In recent years,SERS has been widely used in many fields such as biomedicine,food safety,environmental monitoring,and material analysis.Especially in trace analysis,SERS even allows the detection of single molecules.However,due to the inevitable impact of the instruments and techniques for measuring SERS spectra,the measured raw spectrum contains a large amount of noise,fluorescence background,and interference from other components.At the same time,SERS data is huge in size and contains much redundancy,which may seriously decrease the stability and reliability of Raman analysis.Therefore,how to eliminate the interfere and extract effective information from the complex original spectra quickly and accurately is a key problem to be solved by SERS spectral data processing technology.In order to solve the issue that the ubiquitous fluorescence background of SERS spectroscopy interferes with spectral analysis terribly,a novel algorithm for baseline correction of Raman spectra is proposed in this thesis,which is named as Iteratively Asymmetrically Weighted Penalized Least Squares?IAWPLS?.The method works by iteratively adjusting weights of the difference between the fitted baseline and the original signals,introducing the idea of partially balanced weighting by a softsign weighting function,avoiding the common phenomenon that the final baseline is underestimated in the no peak region and the height of peaks might be overestimated.As a result,the proposed method enables adaptive baseline correction of different SERS spectra and outperforms the existing methods in accuracy,stability,and convenience.In addition,the degree of aggregation of the corrected spectrum is significantly strengthened,indicating that this method successfully removes background signal interference and preserves the peak information of SERS spectra.Due to the high dimensionality and redundancy of the SERS spectrum,the efficiency of spectral analysis is low,and the accuracy and the robustness of model are poor.This thesis studies the spectral feature extraction method.The Successive Projections Algorithm?SPA?is used to extract the spectral wavenumber intervals,which partly solves the problem that individual variables extraction easily causes the loss of important feature information.The method uses the different interval lengths to perform variable selection on the mixture spectra of 4-MBA,R6G and BSA and the mixture spectra of 4-MBA and 4-MPY.Then four classifiers,PCA-LDA,PLS-DA,SIMCA,and KNN,have been constructed to classify.According to the experimental results,this method not only keeps the classification accuracy but also simplifies the classification model and improves the computational efficiency.It is of great significance for online analysis by SERS spectra.