Content area
Abstract
Accurate identification and control of adulteration present a crucial and complex issue in the extra virgin olive oil (EVOO) industry. This study compares fluorescence hyperspectral imaging (FHSI) and FTIR spectroscopy for identifying adulteration in EVOO with various types and concentrations of vegetable oils. FHSI data encompass both fluorescence spectra and images. Five algorithms were employed to establish discrimination models for EVOO adulteration. FTIR, when coupled with a support vector machine, demonstrated the best performance, achieving an accuracy of 99.1% and a detection limit of 2.5%. The classification accuracy of fluorescence spectra combined with a random forest model reached 90.1%, with a detection limit of 5%. Additionally, the optimal classification accuracy of fluorescence images paired with convolutional neural network reached 94.2%, also with a detection limit of 5%. In light of these findings, FTIR is more suitable for high-precision detection, whereas FHSI is better suited for large-scale detection due to its efficient batch processing capabilities.