Espectroscopia no infravermelho próximo e análise multivariada na avaliação do potencial fisiológico de sementes de gergelim e sorgo

Abstract

Investing in faster and more efficient techniques for evaluating seed vigor guarantees the safe and rational allocation of resources and energy, promotes scientific and technological development related to the production of seeds and grains. The combination of spectroscopy with multivariate analysis can improve the evaluation and/or selection of seeds and grains, minimize human analyst error and reduce costs. This study aimed to investigate the potential of near-infrared spectroscopy, combined with multivariate analysis, to evaluate and classify the physiological potential of sesame and sorghum seed lots. The first study was carried out with nine lots of sesame seeds, while the second used four lots of sorghum seeds. The lots were classified into different vigor levels using partial least squares regression (PLS-DA) discriminant analysis, using NIR and chemometric tools. The PLS-DA regression model, with cross-validation, had the best result with the pre-treatments second derivative, second derivative and light scattering correction and Smoothing + first derivative for high, medium and low seed vigor levels, respectively, with more than 80% efficiency in the classification of sesame seeds. For sorghum seeds, the result that stood out the most was with the pre-treatment second derivative + SNV, with 100% efficiency in classification. The main NIR spectra of sorghum and sesame seeds occur in the regions between 997 to 1452 nm and 1040 to 1449 nm, respectively. The interpretation of NIR spectra indicates that water was the most important compound in sorghum seeds; for sesame, the presence of aliphatic hydrocarbons was more pronounced. The external validation tests indicated that sorghum and sesame seeds, with different levels of vigor, were classified with 100% and about 80% of correct answers, respectively. Therefore, the results of this study indicate that the combination of near-infrared spectroscopy with multivariate analysis was effective to evaluate and classify the physiological potential of sesame and sorghum seed lots.