Predicting Mildew Contamination and Shelf-Life of Sunflower Seeds and Soybeans by Fourier Transform Near-Infrared Spectroscopy and Chemometric Data Analysis

Abstract

Expired food products could lead to serious food safety crises without early warnings. Fourier transform near infrared spectroscopy (FT-NIRS) combined with chemometric methods were proposed to discriminate quality (e.g., mildew contamination and quality-guaranteed) and predict shelf life of sunflower seeds and soybeans. For the data analysis, supervised pattern recognition models including principal component analysis (PCA), linear discriminant analysis (LDA), and partial least squares discriminant analysis (PLSDA) were at first constructed to extract variables and identify the differences of NIRS fingerprint information between the aforementioned products in different shelf life stages. Multivariate calibration models based on the partial least squares regression (PLSR) with different spectra preprocessing were subsequently developed to predict their shelf life ranging from 8 to 30 days for soybeans and 30 to 125 days for sunflower seeds, respectively. As a result, compared with the PCA and LDA, optimized PLSDA with reduced complexity was considered as the best model leading to an accurate rate with 100% for nondestructive recognition of the mildewed and quality-guaranteed products in different shelf life. In addition, PLSR based on second-derivative spectra was obtained with the best performance for modeling the shelf life of samples (e.g., RMSEP = 2.35 days for sunflower seeds, RMSEP = 0.61 days for soybeans). In conclusion, FT-NIR spectroscopy and chemometrics have demonstrated the potential for quality control in rapid discrimination and prediction of the quality of sunflower seeds and soybeans, which could also be applied for other food and/or other products.