Rapid non-invasive assessment of quality parameters in ground soybean using near-infrared spectroscopy

Abstract

Abstract: The objective of this work was to evaluate multivariate calibration models to predict total lipids, crude protein, and moisture content in grinded soybean grains using near-infrared spectroscopy and partial least squares (PLS). Three hundred samples of grinded soybean, evaluated in duplicate, were used for reference and spectral measurements. The PLS models for total lipids, crude protein, and moisture were validated by figures of merit for accuracy and precision, respectively, of 0.75 and 0.67 for total lipids, 0.51 and 0.46 for crude protein, and 0.97 and 0.99 for moisture. The PLS models developed for total lipids, crude protein, and moisture can be used as an alternative methodology for the determination of physicochemical parameters, and, therefore, they can be applied in quality control in soybean processing industries.

Keywords

• Glycine max
• chemometric methods
• crude protein
• multivariate calibration
• partial least squares
• total lipids