Hyperspectral Microscopy Technology to Detect Syrups Adulteration of Endemic Guindo Santo and Quillay Honey Using Machine-Learning Tools

Abstract

Honey adulteration is a common practice that affects food quality and sale prices, and certifying the origin of the honey using non-destructive methods is critical. Guindo Santo and Quillay are fundamental for the honey production of Biobío and the Ñuble region in Chile. Furthermore, Guindo Santo only exists in this area of the world. Therefore, certifying honey of this species is crucial for beekeeper communities—mostly natives—to give them advantages and competitiveness in the global market. To solve this necessity, we present a system for detecting adulterated endemic honey that combines different artificial intelligence networks with a confocal optical microscope and a tunable optical filter for hyperspectral data acquisition. Honey samples artificially adulterated with syrups at concentrations undetectable to the naked eye were used for validating different artificial intelligence models. Comparing Linear discriminant analysis (LDA), Support vector machine (SVM), and Neural Network (NN), we reach the best average accuracy value with SVM of 93% for all classes in both kinds of honey. We hope these results will be the starting point of a method for honey certification in Chile in an automated way and with high precision.

Classification of Monofloral Honeys by Measuring a Low-Cost Electronic Nose Prototype Based on Resistive Metal Oxide Sensors

“…In the second experiment, the machine learning (ML) algorithm was used through Support Vector Machines (SVM). This method has given good results in other recent studies related to honey adulteration or classification that have combined machine learning tools [32,33] with other analysis techniques such as isotope profiles, hyperspectral microscopy technology, or infrared spectroscopy [34].…”

Rapid and Automated Method for Detecting and Quantifying Adulterations in High-Quality Honey Using Vis-NIRs in Combination with Machine Learning

“…Additionally, the chromatographic methods need several chemicals and usually produce waste. Therefore, other methodologies have been used in recent years to detect various types of adulteration in honey, including ion mobility spectrometry (IMS) [ 16 , 17 , 18 ], nuclear magnetic resonance (NMR) [ 19 , 20 ], DNA-based techniques [ 21 , 22 ], and pollen visualization [ 23 , 24 ], among others. Among them, pollen visualization stands out, as it is a fast and cheap methodology that can be combined with machine learning algorithms, allowing for the characterization of honeys in an objective way [ 23 ].…”