Abstract:

In order to reduce computational complexity of modeling and improve the model's robustness and prediction accuracy, successive projections algorithm (SPA) was used in the near infrared spectrum calibration modeling of wheat gluten. Firstly, a representative set of correction samples were selected by SPXY algorithm. Secondly, the spectral data was pretreated with several different methods to enhance spectral features. Thirdly, making use of SPA to extract sensitive wave points of the original spectrum and the spectrum after preprocessing and then multiple linear regression (MLR) calibration models were established. The results showed that the calibration model established with the data extracted from the spectrum after standard normal variate transformation (SNV) obtained the best results. The root mean square error of prediction (RMSEP) and the prediction correlation coefficient (r) were 1.3332 and 0.94319, respectively, which was better than the model established by partial least square regression (PLSR) under the same conditions.