Differentiation of sunflower cultivars by spectroscopy in infrared next and multivariate analysis using seeds and oil

Abstract

This work aimed to evaluate the near-infrared spectroscopy technique (NIR) and multivariate analysis in the differentiation of sunflower cultivars, using seeds and oil. The samples were subjected to analysis in the NIR and the spectra were generated by the FT-IR detector. To construct the calibration model it was used the multivariate classification method of partial least squares-discriminant analysis (PLS-DA), in which the classes (y) are the dependent variables and the samples’ spectra are the independent variables. Sunflower cultivars were differentiated both by oil and by seed. For oil it was obtained 100% accuracy in the calibration, 92% in y-randomization test, 86% in cross-validation and 92% in external validation in which 25% of samples are tested to validate the model. And seeds had 100% accuracy in the calibration, 87% in y-randomization test, 100% in cross-validation and 100% in external validation. Therefore, it is concluded that the near-infrared spectroscopy associated with multivariate analysis differentiates sunflower cultivars, both by oil (extracted from seeds with and without pericarp) and by seed (with and without pericarp).