Protein and Oil Contents Determination in Wheat Using Near-infrared (NIR) Hyperspectral Imaging

Abstract

Grain quality laboratories around world strive for an effective technique to determine protein and oil contents in wheat. Feasibility of near-infrared (NIR) hyperspectral imaging technique was assessed by developing prediction models to determine protein and oil contents of common wheat classes grown in western Canada. Wheat bulk samples were scanned in a wavelength region of 960 1700 nm at 10 nm intervals using a long wavelength indium gallium arsenide (InGaAs) NIR camera. Seventy five NIR absorbance intensities were extracted from the scanned images and used for developing prediction models for protein and oil contents of wheat using the partial least squares regression (PLSR) method. Twelve and thirteen partial least squares (PLS) factors were used to develop PLSR models to predict protein and oil contents, respectively. The developed models explained 89% of protein variation and 68% of oil content variation, respectively, in wheat. Correlation coefficients of 0.94 and 0.83 were obtained for predicting protein and oil contents, respectively, using PLSR models. These results establish that NIR hyperspectral imaging can be used as an effective method to determine protein and oil contents in wheat.