AI-based hyperspectral and VOCs assessment approach to identify adulterated extra virgin olive oil
Author(s)
Date Issued
2021
Type
article
Volume
247
Issue
4
Start Page
1013
End Page
1022
Abstract
Spectral and VOCs data are important to determine the genuineness of the EVOO. The aim of this work is to apply an open-source device for VIS–NIR spectrophotometric measurements and a VOCs analyzer to identify some types of sophistication. 96 samples were analyzed. Samples included pure EVOO compound olive oil pure seed oils and olive oil samples adulterated with 7 different seed oils at different ratios. An artificial intelligence model was applied to identify adulterations from the spectral and VOCs data. The models built on both spectral and VOCs data showed perfect classification of pure EVOO samples with respect to sophisticated ones in both model and test sets. The most important spectral values and VOCs were extracted. Artificial intelligence applicative models aimed not only to identify sophisticated samples but also to understand the most informative spectra and VOCs to prototype specific devices for anti-fraud control.
Additional information
L.O. implemented the machine learning algorithm for the spectral analysis of EVO and designed and implemented the algorithm for the analysis of the feature importance
Subjects
Project(s)
MiPAAF-INFOLIVA (D.M. n. 12479)