Rapid and Non-destructive Detecting Frying Times of Peanut Oil Based on Near Infrared Reflectance Spectroscopy

Aiming at the food safety problem of edible oil, this paper proposes a new method based on Near Infrared Reflectance Spectroscopy (NIRS) for rapid non-destructive testing of the frying times of frying oils. Peanut oil is used as a research object, and frozen French fries are used as a frying medium. Peanut oil is subjected to 10 experiments, and each experiment is fried 15 times in the same batch, and the samples are collected in the near-infrared original spectra at 400 nm-2500 nm. The original spectra are preprocessed and combined with the data dimensionality reduction algorithm to establish the classification model and the regression model of frying times of peanut oil, and the accuracy of the model prediction is tested. Choose the first derivative as the pretreatment method and the Linear Discriminant Analysis (LDA) algorithm is used to reduce the dimensionality of the preprocessed spectral data to establish a K-Nearest-Neighbors (KNN) classification model for peanut oil. The prediction effect of the Random Forest Regression (RFR) regression model based on the spectral data after dimensionality reduction is slightly better than that of the Partial Least Squares Regression (PLSR) regression model. The Determination Coefficient (R2), Root Means Square Error (RMSEP), and Relative Analysis Error (RPD) of the peanut oil in RFR regression models are 0.9978, 0.1823, 21.2776. Therefore, the method used in this study can effectively detect the frying times of peanut oil and provide a technical guarantee for the rapid detection of food safety.