Identification and quantification of counterfeit sesame oil by 3D fluorescence spectroscopy and convolutional neural network. (1st May 2020)

• Record Type: Journal Article
• Authors: Wu, Xijun; Zhao, Zhilei; Tian, Ruiling; Shang, Zhencheng; Liu, Hailong
• Is Part Of: Food chemistry. Volume 311(2020)
• Publication Date: 2020-05-01
• Subjects: Identification of counterfeit -- 3D fluorescence spectrum -- Convolutional neural network -- Support vector machine -- Partial least squares; Food -- Analysis -- Periodicals; Food -- Composition -- Periodicals; 664
• DOI: 10.1016/j.foodchem.2019.125882
• Languages: English
• ISSNs: 0308-8146

Abstract

Highlights: Vegetable oil samples were characterized by 3D fluorescence spectroscopy. Convolutional neural network was transferred to extract spectral characteristics. Support Vector Machine determined the counterfeit sesame oil and its ingredients. Partial least squares successfully predicted the level of counterfeit sesame oil. The preprocessing of image eliminated the need for extensive experimental samples. Abstract: The method of 3D fluorescence spectroscopy combined with convolutional neural network (CNN) was developed to identify the counterfeit sesame oil. AlexNet, a pre-trained CNN architecture, was transferred to extract spectral characteristics. Then these features extracted by AlexNet were used as the input of the support vector machine (SVM) to determine whether the sample was counterfeit and its ingredients simultaneously, and both the accuracy were 100%. According to different counterfeit ingredients, these features extracted by AlexNet were used as the input of partial least squares (PLS) to predict the volume percentage concentration of sesame oil essence. There was a good linear relationship between the predicted and actual values of the three sets of counterfeit samples (R 2 > 0.99), and the root mean square error of prediction (RMSEP) values were 0.99%, 2.20% and 1.64%, respectively. The results confirmed the validity of this novel method in sesame oil identification.