Development and Optimization of a Rapid Detection Model for Characteristic Nutrient Content in Highland Barley

In this paper, the highland barley moisture, β-glucan, amylopectin and protein content rapid detection model was established based on the 4000~10000 cm⁻¹ near-infrared spectrum and actual contents of 76 highland barleys. The results showed that SG convolution smoothing was the optimal spectral preprocessing method for the partial least squares (PLS) prediction model of moisture, amylopectin and β-glucan contents, while SG convolution smoothing+multiplicative scatter correction (MSC) was the optimal spectral preprocessing method for the PLS prediction model of protein content. In order to further improve the accuracy of the prediction model, the different characteristic wavelength selection algorithms including competitive adaptive reweighted sampling (CARS), successive projections algorithm (SPA) and variables combination population analysis and iterative retained information variable (VCPA-IRIV) were used on the prediction results of the model. The results showed that VCPA-IRIV treatment could effectively improve the determination coefficient of prediction of moisture, amylose and protein content prediction model, and reduce the root mean square error of prediction. The treatment of CARS had a remarkable effect on the prediction accuracy for the β-glucan content. Ultimately, the established prediction models of moisture, β-glucan, amylopectin and protein content for highland barley had good prediction accuracy with the appropriate R_p (0.9868, 0.9808, 0.9701 and 0.9879) and RMSEP (0.2042, 0.1846, 0.8135 and 0.2095) value, respectively. In conclusion, the rapid detection model of highland barley characteristic nutrient content based on near infrared spectroscopy established in this study had high accuracy, which would have certain guiding significance for processing enterprises to quickly understand the quality of raw materials and efficiently screen qualified raw materials.