Comparison of NIR and Raman spectra combined with chemometrics for the classification and quantification of mung beans (Vigna radiata L.) of different origins

In this study, we compared two technologies (i.e. Near-infrared and Raman spectroscopy) for origin identification and quantitative research on nutritional components of mung beans based on the chemometric principles. The orthogonal partial least squares discriminant analysis models with Near-infrared as well as Raman spectroscopy had a predictive ability to 94.3% and 92.9%, respectively, indicating that differentiation of mung beans from different origin sources could be achieved by both Near-infrared and Raman spectroscopy. Quantitative models for moisture, protein and total starch were performed using partial least squares regression techniques based on different spectral pre-processing methods. Overall, the partial least squares quantitative regression model built with Near-infrared showed better performance than that of Raman spectroscopy. The partial least squares regression model obtained by multiplicative scatter correction combined with first derivative treatment of Near-infrared spectral data showed excellent predictive ability (Rc = 99.9%, Rp = 85.3%) for moisture. The quantitative protein prediction model built by multiplicative scatter correction treatment of Near-infrared also performed well (Rc = 91.4%, Rp = 91.5%). In addition, we also characterized potential differential compounds in mung beans of different origins by UPLC-Q-TOF-MS. This study provides a theoretical basis for the traceability of legume products and the construction of multiple rapid detection methods.