Research On Application Technology Of Raman Spectroscopy In Qualitative Identification And Classification Of Gemstone

Raman spectroscopy is a fast and nondestructive analytical detection technique,and Raman spectrum contains a wealth of material molecular structure information.Based on the application status of Raman spectroscopy in gemstone field,this paper further explored its application technology in the field of qualitative identification and origin distinction of gemstone.This paper integrated a miniaturized Raman probe,through which the laser and the outer light path probe were integrated to build a miniaturized Raman spectroscopy test system with a spectral resolution of 6-8cm-1 and a spectral range of 200-2700cm-1;Participated in the design of the spectrum acquisition and processing software based on the Android platform,including four functional modules of spectral acquisition,spectral data preservation,spectral comparison library and parameter settings.Aiming at the gemstone field researched in this paper,a gemstone identification software was designed and developed independently,which contained three functions of spectral information storage,query and spectrum discrimination.This paper achieved the identification and differentiation of emerald from different origins based on Raman spectroscopy and principal component analysis.Firstly,Raman spectra of emerald samples from known origins were collected,based on the mean Raman spectra of emeralds,the Raman characteristic peaks that can be used as the basis for distinguishing emeralds were extracted,and the principal components analysis was used to distinguish the emeralds of Zambia,Colombia and Brazil,the accuracy was 100%.In this paper,a new method based on Raman spectroscopy and pattern recognition algorithm for qualitative identification and origin distinction of nephrite was proposed.By using the Raman specscropy measurement system,the spectra of nephrite samples from the known origin were obtained,the collected Raman spectra were divided into training set and test set,and the training set were modeled by KNN algorithm,PCA-LDA and SVM,and then the models were validated by the test set.Through the comparison of the verification results,it was found that KNN algorithm and PCA-LDA can realize distinction of nephrite from two different origins,but the classification accuracy of PCA-LDA is better;On this basis,the model established by SVM had the accuracy of nearly 100% for distinguishing samples from two different origins;the accuracy of simultaneously distinguishing the three or four kinds of nephrite was more than 90%.Up to five origin nephrite samples were simultaneously discriminated during the experiment.In this paper,a rapid and non-destructive method based on Raman spectroscopy combined with pattern recognition algorithm for discriminating the origin of Tianhuang was proposed.Raman spectroscopy data were collected from Tianhuang samples of three known origins and divided into training set and test set.Based on Mahalanobis distance discriminant method,PCA-LDA and Random forest algorithm,the models of training set were built,and then the corresponding models were validated by test set.The validation results showed that Mahalanobis distance discriminant method had a low accuracy of discriminating the origin of Tianhuang,only about 77%;PCA-LDA had a classfication accuracy more than 90% for the two origins of Tianhuang,but it could not achieve the simultaneous discrimination of the Tianhuang more than two origins;Compared to the previous two algorithms,Random forest can achieve the discrimination of two origins of Tianhuang with the accuracy more than 90%,and the distinction accuracy of simultaneously discriminated three origins of Tianhuang was 80%.