Fourier Transform Near-Infrared Spectroscopy to Predict the Gross Energy Content of Food Grade Legumes

Abstract

The feasibility of Fourier transform near-infrared reflectance spectroscopy (FT-NIRS) for determining gross energy content in different food legumes has been investigated. Eighty food-grade legume samples were obtained from different retailers and local markets in Hungary and they included 42 common beans (Phaseolus vulgaris L.), 20 peas (Pisum sativum L.), 10 lentils (Lens culinaris L.), and 8 soya beans (Glycine max L.) both as full fat food and defatted. The samples were analyzed by an adiabatic bomb calorimeter and then scanned in a Bruker MPA FT-NIR Analyzer (800–2,500 nm). Two algorithms for spectral selection of calibration and validation samples, which represent variability encountered in the full population, were tested. Partial least squares regression were developed for the prediction of gross energy using four different spectral preprocessing methods (first and second derivative alone and combined with standard normal variation and multiplicative scatter correction). The results show that first derivative produced the most accurate results with very high coefficient of determinations in validation (<93 %) and with very low standard errors of validation (<0.025 kcal/g) as compared to the standard error of the reference method (0.204 kcal/g).