Articles
Non-invasive wine authentication method using near-infrared spectroscopy through the bottle
Article number
1387_16
Pages
121 – 126
Language
English
Abstract
Wine fraud is currently a massive problem within the wine industry, with counterfeiting and adulteration the most frequent practices, which have led the industry to lose billions of dollars annually.
Hence, the importance of developing rapid, accurate, and non-invasive techniques to identify signs of fraud, such as wine provenance and quality traits.
This study aimed to develop a non-invasive technique using a hand-held near-infrared (NIR) spectroscopy device (1596-2396 nm) to measure wines through the bottle.
Machine learning models based on artificial neural networks were developed using the NIR absorbance values as inputs to predict the wine vintage (classification Model 1) and intensity of sensory descriptors (regression Model 2) of Australian Shiraz wine.
Models resulted in high accuracy, with 97% for Model 1 and R=0.95 for Model 2. This is an effective, affordable, and rapid method for winemakers and retailers to assess the wines without the need to open the bottles and may be further trained to assess other authentication indicators such as region and country of origin, and grape cultivar.
Hence, the importance of developing rapid, accurate, and non-invasive techniques to identify signs of fraud, such as wine provenance and quality traits.
This study aimed to develop a non-invasive technique using a hand-held near-infrared (NIR) spectroscopy device (1596-2396 nm) to measure wines through the bottle.
Machine learning models based on artificial neural networks were developed using the NIR absorbance values as inputs to predict the wine vintage (classification Model 1) and intensity of sensory descriptors (regression Model 2) of Australian Shiraz wine.
Models resulted in high accuracy, with 97% for Model 1 and R=0.95 for Model 2. This is an effective, affordable, and rapid method for winemakers and retailers to assess the wines without the need to open the bottles and may be further trained to assess other authentication indicators such as region and country of origin, and grape cultivar.
Publication
Authors
C. Gonzalez Viejo, N. Harris, C. Barnes, E. Tongson, C. Hernandez-Brenes, J. Valiente-Banuet, S. Fuentes
Keywords
machine learning, wine fraud, wine quality, rapid methods, artificial intelligence
Online Articles (46)