Near-infrared-based Identification of Sesame Oil Authenticity

[摘要] Near-infrared spectroscopy (NIRS) combined with chemometrics were employed to determine the quality of sesame oil. Sesame oil authenticity identification and content determination models with two variables (binary systems) were established. 52 sesame oil samples were mixed with soybean and rapeseed oil for the preparation of fake sesame oil samples. NIRS, in association with the support vector machine classification (SVC), was used to establish full-range-wavelength models. The competitive adaptive reweighted sampling (CARS) and backward interval partial least squares (BIPLS) methods were adopted in the optimization process for characteristic wavelengths for the modeling. The results showed that all the models established could effectively identify the authenticity of sesame oil. The data were preprocessed with the standard normal variable transformation algorithm (SNV), and the preprocessed data were then used to establish an SNV-SVC model with the prediction set recognition accuracy of up to 99.4975%. The correlation coefficient R of the prediction model for content determination was higher than 99%, and the mean square error (MSE) was lower than 0.0605, indicating that the models based on NIRS and support vector machine regression (SVR) can realize the content determination of fake sesame oil.

[发布机构] School of Mechanical Engineering, Wuhan Polytechnic University, China^1

[效力级别] 无线电电子学 [学科分类] 材料科学（综合）

[关键词] Authenticity identifications;Content determination;Correlation coefficient;Interval partial least squares;Recognition accuracy;Support vector machine classification;Support vector machine regression (SVR);Variable transformation