Abstract:
To establish a rapid method for discriminating among sesame oil processing methods, this study investigated stone-milled, pressed, and refined sesame oils. The fatty acid compositions and near-infrared (NIR) spectral characteristics of sesame oils produced by these three methods were systematically analyzed. Discrimination models were developed using NIR spectroscopy combined with principal component analysis-linear discriminant analysis (PCA-LDA), PLS-DA, and SVM. The results indicated that the processing method had a relatively minor effect on the fatty acid composition and content of the sesame oils. However, the NIR spectral differences between refined sesame oil and the other two types (stone-milled and pressed) were substantial. Among the three models, the SVM model exhibited the best performance, achieving a classification accuracy of 100%. All models demonstrated higher identification accuracy for refined sesame oil than for stone-milled and pressed sesame oils. For stone-milled sesame oil, the sensitivity of the three methods followed the order: SVM (100.0%) > PCALDA (87.5%) > PLS-DA (75.0%). For pressed sesame oil, the sensitivity ranking was SVM (100.0%) > PLS-DA (83.3%) > PCA-LDA (66.7%). These results demonstrate that NIR spectroscopy coupled with SVM can effectively discriminate sesame oil processing methods, providing a technical reference for oil authenticity testing.