Study On Identification Technology Of Vegetable Oil Based On Raman Spectroscopy And Convolutional Neural Network

Blended vegetable oil is a kind of vegetable oil mixed by two or more pure vegetable oils in a certain proportion.Because it has more balanced nutrients and is loved by people,it has become an important product in the vegetable oil market.Therefore,it is of great practical significance to identify the quality of blended vegetable oil.Raman spectroscopy has been widely concerned by researchers because of its non-destructive,rapid and high sensitivity.It has become an important detection method in the field of food identification.The rapid development of deep learning provides a new method to solve the multicollinearity problem of spectrum.In this study,Raman spectroscopy combined with convolutional neural network(CNN)was used to qualitatively and quantitatively identify blended vegetable oil.The main research work is as follows:(1)Taking corn oil(CO),peanut oil(PO)and extra virgin olive oil(EVOO)as the research objectives,the blended vegetable oil prepared from the mixture of three pure vegetable oils was characterized by Raman spectroscopy,and the relationship between the content of EVOO in blended vegetable oil and the intensity of Raman characteristic peak was explored.(2)Raman spectroscopy combined with CNN was used to qualitatively identify three kinds of blended vegetable oils containing EVOO.The Raman spectra of blended corn-olive oil(CO-EVOO),blended peanut-olive oil(PO-EVOO)and blended corn-peanut-olive oil(CO-PO-EVOO)samples were collected by confocal Raman spectrometer to construct the datasets of three kinds of blended vegetable oils.CNN classification model was established to qualitatively identify the samples without EVOO,the samples with low content of EVOO and the samples with high content of EVOO in three kinds of blended vegetable oil samples,and compared with the traditional classification and identification methods support vector machine(SVM)and partial least squares discriminant analysis(PLS-DA)to evaluate the performance of CNN classification model.(3)The EVOO in CO-EVOO was quantitatively identified by Raman spectroscopy combined with CNN.The Raman spectra of CO-EVOO samples were collected by confocal Raman spectrometer,and a dataset with 315 Raman spectra in 7 proportions was constructed.The CNN regression analysis model was established to quantitatively identify EVOO in CO-EVOO samples.Compared it with support vector regression(SVR)and partial least squares regression(PLSR)to evaluate the performance of CNN regression analysis model.The results show that the classification and regression model based on Raman spectroscopy and CNN has great application potential in the qualitative and quantitative identification of blended vegetable oil.