Using UV-induced fluorescence images integrated with a CACHAS-based two-step hierarchical classification approach for rapid detection of extra virgin olive oil adulteration

Abstract

This study introduces a novel environmentally-friendly two-step hierarchical classification approach combined with UV-induced fluorescence images for authenticating extra virgin olive oil (EVOO) and detects individual and simultaneous adulteration with refined soybean, canola, and sunflower oils. A Chemometrics-Assisted Color Histogram-based Analytical System (CACHAS) was developed to capture digital fluorescence images and extract analytical fingerprints in the Grayscale, RGB, and HIS channels. Data-Driven Soft Independent Modeling of Class Analogy (DD-SIMCA), optimized for one-class classification, was employed to authenticate pure EVOO with an efficiency of 99.8 % samples, while Partial Least Squares Discriminant Analysis (PLS-DA), designed for multi-class discrimination, was utilized to discriminate the adulterated samples based on the presence of one, two, or three refined vegetable oils with an accuracy of 95.6 %. The complementary application of these methods allows a robust evaluation of the dataset, with DD-SIMCA focusing on validating authentic samples and PLS-DA enhancing the discrimination of adulterated samples across multiple categories.