Rapid and non-destructive determination of protein and starch content in agricultural powders using near-infrared and fluorescence spectroscopy, and data fusion

Triboelectrostatic separation is a promising technology for the fractionation of agricultural flours to produce protein- and starch-enriched fractions. Rapid determination of the protein and starch content of these powder fractions is required for process development, optimization and ultimately control. The suitability of near-infrared and fluorescence spectroscopy along with data fusion for the rapid determination of protein and starch content and the development of a robust multivariate regression model was the objective in this work. A total of 102 samples were collected following triboelectrostatic separation of bean flour and analyzed using NIR and fluorescence spectroscopy. The superior performance of partial least squares method for both protein and starch, confirmed that NIR spectroscopy is a reliable method for rapid determination of protein- and starch-rich powders. The NIR and fluorescence data fusion model performed with the highest accuracy to predict the protein content.