Veuillez utiliser cette adresse pour citer ce document :
Titre: On-line system based on hyperspectral information to estimate acidity, moisture and peroxides in olive oil samples
Auteur(s): Martínez Gila, Diego Manuel
Cano Marchal, Pablo
Gámez García, Javier
Gómez Ortega, Juan
Résumé: The analysis of the quality of virgin olive oil involves the determination of a series of properties, such as chemical indexes and organoleptic characteristics. In addition, the determination of these properties in real-time could be useful in order to improve the olive oil extraction process since the process parameters could be regulated with the real-time moisture information. In this paper, the feasibility of using a non-invasive hyperspectral device, in order to determine on-line three parameters of the olive oil (free acidity, peroxide index and moisture) is studied. In order to study the correlation between these parameters and the information obtained by the hyperspectral sensor (absorption level), three different methods were applied: genetic algorithms (GA), least absolute shrinkage and selection operator (LASSO), and successive projection algorithm (SPA). From the experimental results, reduced values in cross validation were obtained and the optimal wavelengths were pointed out.
Mots-clés: Olive oil
Process control
Hyperspectral imaging
Machine learning
Date de publication: 18-jui-2015
metadata.dc.description.sponsorship: DPI2011-27284, TEP2009-5363 and AGR-6429.
Référence bibliographique: Martínez Gila, D., Cano Marchal, P., Gámez García, J., & Gómez Ortega, J. (2015). On-line system based on hyperspectral information to estimate acidity, moisture and peroxides in olive oil samples. Computers and Electronics in Agriculture, 116, 1–7.
Collection(s) :DIEA-Artículos

Fichier(s) constituant ce document :
Fichier Description TailleFormat 
On-line system based on hyperspectral_ACCEPTED_VERSION.pdf1,25 MBAdobe PDFVoir/Ouvrir

Ce document est protégé par copyright

Ce document est autorisé sous une licence de type Licence Creative Commons
Creative Commons