A non-destructive determination of protein content in potato flour noodles using near-infrared hyperspectral imaging technology

Abstract

Hyperspectral imaging (HSI) with the near-infrared (NIR) bands (900–2250 nm) was employed to investigate the non-destructive prediction of protein content in potato flour noodles. The protein content (8.906–10.515 %) of 120 potato flour noodles was studied to establish the prediction model. Partial least squares regression (PLSR) model was established based on the spectra of potato flour noodles to predict the protein content showing high performance. After optimization, orthogonal signal correction (OSC) was used to preprocess the original spectra, and the competitive adaptive reweighted sampling algorithm (CARS) was chosen to select characteristic wavelengths, therefore, OSC-CARS-PLSR was established. Next, 77 samples were selected as the calibration set, and the remaining 38 samples were used as the prediction set. The coefficient of determination (R2) and the root mean square error (RMSE) were used to evaluate the performance of the model. OSC-CARS-PLSR showed a high performance with R2 values of 0.9606 and 0.8925 and RMSE values of 0.070 % and 0.1385 % in the calibration set and prediction set, respectively. The visualization image was used to identify protein distribution in potato flour noodles. Overall, the results indicate that HSI technology could accurately predict the protein content in potato flour noodles providing a rapid and non-destructive method to detect protein and other compositions in grains and foods.