A strategy to improve the detection accuracy and universality of highland barley muti-quality attributes based on near infrared spectroscopy combined with model transfer method.

This study utilized direct standardization (DS), piecewise direct standardization (PDS), and DS-PDS algorithms to facilitate model transfer across instruments during the detection of highland barley (HB), and to calibrate HB samples in various states. Following the model transfer, the slave instrument results showed significant improvement. After DS processing, optimal results were achieved for total starch (prediction set correlation coefficient, R = 0.903), amylose (R = 0.936), β-glucan (R = 0.929). After DS-PDS processing, the best results were observed for protein (R = 0.855) and total phenols (R = 0.937). Moreover, after scatter correction, the results for grains were further enhanced. Following DS treatment, total starch (R = 0.873), amylose (R = 0.950), protein (R = 0.899), β-glucan (R = 0.899), and total phenols (R = 0.965) demonstrated optimal performance. Overall, this study significantly enhanced the universality and accuracy of models through effective model transfer and scatter correction for grains.