Research On Detection Of Camellia Oil Adulteration Based On Raman Spectroscopy

Camellia oil has rich nutritional value and is loved by more and more people,but the adulteration of camellia oil is also becoming more and more serious.Therefore,we urgently need to find a technology that can quickly and accurately detect the adulteration of camellia oil,which is of great significance to the quality assurance of camellia oil and people’s health.This paper mainly studies Raman spectroscopy combined with LDA or other analysis methods for adulteration detection of camellia oil.In this paper,after adding corn oil,rice oil and soybean oil into camellia oil at different concentrations,binary adulterated samples of camellia oil and ternary adulterated samples of camellia oil were obtained,and all samples were tested by Raman spectroscopy.Firstly,the spectra of the samples were preprocessed to reduce the effects of factors such as noise.Then,the Raman spectra of the samples were analyzed,and it was found that the peak intensity of some characteristic peaks of different oils were significantly different,which provided the possibility for us to identify adulteration of camellia oil by Raman spectroscopy.In terms of qualitative analysis,three binary adulteration models of camellia oil were established by linear discriminant analysis based on principal component analysis(PCA-LDA),and all samples were classified correct;four kinds of pure oils and three kinds of adulterated oils were simultaneously identified by the support vector machine(SVM).The classification accuracy of the training set was 99.6%,and that of the prediction set was 100%.In terms of quantitative analysis,a BP-Adaboost neural network model based on Raman spectrum is proposed to predict the concentration of soybean oil in adulterated camellia oil.The BP-Adaboost neural network model is constructed by spectral characteristic variables extracted by PCA,and the prediction results of the model were compared with the other three regression methods.The results show that the BPAdaboost neural network has the best performance in predicting adulteration concentration.The correlation coefficient of the training set is 0.998 and the root mean square error is 1.127%,while the correlation coefficient of the prediction set is 0.999,the root mean square error is 0.697%.In qualitative analysis of ternary adulteration of camellia oil,the PCA-LDA method was used to successfully distinguish pure oils and adulterated oils.In the quantitative analysis of ternary adulterated oil,the optimal pretreatment method was determined with different variables(corn oil concentration,rice oil concentration,total adulteration concentration)as dependent variables,and the regression model was established by the partial least squares regression(PLSR).When corn oil concentration,rice oil concentration and total adulteration concentration were used as dependent variables,the predicted correlation coefficients of the regression models were 0.978,0.967,and 0.983,respectively,and the root mean square errors were 0.0257,0.038,and0.0257,respectively.Finally,through the above analysis results,it is feasible to detect adulteration of camellia oil by Raman spectroscopy combined with linear decision analysis method or other data analysis methods.