Abstract:

To explore a rapid determination method of protein content in grains of Panicum miliaceum L., the predicted models for quantitative analysis of protein contents in the grains was built using near infrared transmittance spectroscopy (NITS), and the influence of optics derivative treatment methods combined with different regression methods on the model was studied. The results showed that the predicting effects of the four combinations between the partial least squares (PLS), the modified partial least squares (MPLS) regression method and first derivative and second derivative were similar. MPLS in combination with first derivative was the best model. In this model, the correlation coefficient (1-VR) was 0.8570, the average determination coefficient of validation (R2) was 0.8806, the square error of cross (SEC) was 0.3424, the square error of cross validation (SECV) was 0.3751, and the standard error of prediction (SEP) was 0.454. The NITS model of intact Panicum miliaceum L. grain for protein content determination could be substituted for traditional chemical method.