A model for the detection of β-glucan content in oat grain based on near infrared spectroscopy

Abstract

Oat β-glucan, recognized, a key functional component of oats, presents challenges in its determination due to low efficiency, cumbersome operations, and high costs, thereby posing obstacles to the large-scale production and industrialization of oats. This study aimed to clarify the quantitative relationship between β-glucan content and near-infrared (NIR) spectra in oat grains. Utilizing 63 oat grain varieties as test materials, this study analyzed their β-glucan content and NIR spectra. The focus of the research lay in evaluating the impact of various preprocessing and modeling methods on the accuracy of NIR spectroscopy for detecting β-glucan content in oat grains. The results indicated that the successive projections algorithm-multiple linear regression (SPA-MLR) model with square preprocessing, along with the partial least squares regression (PLSR) and support vector machine regression (SVR) models under multiplicative scattering Correction (MSC) preprocessing, achieved high accuracy, with correlation coefficients (R2) of 0.84, 0.66, and 0.76, respectively. The SPA-MLR model, employing square preprocessing, exhibited optimal performance, with an R2 of 0.84, root mean square error (RMSE) of 0.55, mean absolute error (MAE) of 0.46, and a ratio of performance to standard deviation (RPD) of 2.45, proving effective in accurately determining the β-glucan content in oat grains.