Unveiling the potential: Harnessing spectral technologies for enhanced protein and gluten content prediction in wheat grains and flour

Abstract

Protein and gluten content is one of the most crucial quality characteristics in the wheat industry. However, these properties are measured after grinding wheat kernels into the flour. In this study, grain samples from 38 different wheat cultivars were collected, and their protein, wet and dry gluten content were measured traditionally. Spectral information was obtained using three non-destructive instruments, including benchtop visible-near infrared hyperspectral imaging (HSI), portable short wavelength infrared HSI and Fourier-Transform near-infrared spectroscopy from both whole grains and their flour samples. Partial least squares regression (PLSR) and Gaussian process regression (GPR) with three spectral pre-treatments were used to compare performances and Neighborhood Component Analysis was applied for wavelength selection.

Through HSI, wheat kernels revealed their protein and gluten content with remarkable precision, achieving R²_P values exceeding 0.97 using GPR based on whole kernel data utilising four wavelengths in the Visible range. The key novelty of this work is that it demonstrates the suitability of visible range hyperspectral imaging for direct prediction of protein and gluten with high accuracy, without the need for sample grinding, thus underscoring the significance of visible spectral information in determining protein and gluten-related parameters.