Objective:To study the modeling methods of near infrared spectroscopy(NIR)and mid infrared spectroscopy(MIR)for the rapid detection of protein in infant formula milk powder.Method:The spectral scanning conditions of the milk powder samples were determined,and modeling algorithms were selected.Pre processing methods such as baseline correction,standard normal variate transform,multivariate scattering correction,and first derivative were studied.The reliability of the model was evaluated using interactive validation and external validation.Result:The correlation coefficient(r)of the near-infrared model after standard normal variate transform preprocessing was 0.906,and standard error of cross validation(SECV)was 0.11.This model was used to validate 15 samples in the test set,and the correlation coefficient between the measured and predicted values of the samples was 0.947;the correlation coefficient(r)and SECV of the mid infrared model were 0.633 and 0.15.The correlation coefficient between the measured and predicted values of 15 samples in the test set was 0.685.Conclusion:The near infrared spectroscopy modeling method is accurate,fast,and simple,and can be used for rapid determination of protein content in infant formula milk powder.The mid infrared spectroscopy modeling method is not suitable for rapid determination of protein content in infant formula milk powder.
关键词
近红外光谱法/中红外光谱法/婴儿配方乳粉/蛋白质
Key words
near infrared spectroscopy/mid infrared spectroscopy/infant formula milk powder/protein
维普
万方数据