Feasibility analysis of nondestructive detection of multiple parameters of highland barley by near-infrared spectroscopy.

BACKGROUND

Highland barley is widely regarded as a premium cereal grain due to its exceptional nutritional profile. This study employed near-infrared spectroscopy technology for the quantitative assessment of five critical parameters in highland barley: total starch, amylose, protein, β-glucan, and total phenols. To optimize spectral data processing, the most effective preprocessing method was identified among six options (standard normal transformation, multivariate scattering correction, normalization (Nor), detrend (DE), first derivative (FD), second derivative (SD)). Furthermore, feature wavelength selection algorithms, including competitive adaptive reweighted sampling (CARS), successive projections algorithm (SPA), uninformative variable elimination, and least angle regression, were utilized to enhance the model's predictive accuracy.

RESULTS

The commendable predictability for total starch was achieved through DE-SPA (R  = 0.913, root mean square error of prediction (RMSEP) = 1.612). For amylose, Nor-CARS exhibited predictive performance (R  = 0.925, RMSEP = 2.049). Protein showcased a creditable result by SD-SPA (R  = 0.876, RMSEP = 0.710). β-Glucan achieved notable predictability through FD-CARS (R  = 0.763, RMSEP = 0.328). Total phenols exhibited remarkable predictability using SD-SPA (R  = 0.946, RMSEP = 0.130).

CONCLUSION

Thus, the study provided a rapid and nondestructive method for monitoring multi-quality parameters of highland barley. © 2025 Society of Chemical Industry.