Real-time detection of turmeric adulteration with metanil yellow using a miniaturized NIR sensor and AI techniques

Abstract

Spices have been among the most targeted foods in the European Union (EU) for fraudsters, given that the spice market exemplifies complex and globalized supply chains. Turmeric is a widely used spice famous for its vivid color, flavor, and purported health advantages. Its medicinal properties in addressing various health issues have sparked a surge in global demand, raising concerns about the spice industry’s integrity. The most common and hazardous adulterants of turmeric, added for financial gain, are synthetic, non-authorized azo dyes, particularly Metanil Yellow (MY). To tackle malpractices in the turmeric supply chain, this study proposes a scalable method utilizing a miniaturized near-infrared (NIR) (1350–2500 nm) sensor coupled with advanced Artificial Intelligence (AI) techniques to detect the presence of MY in turmeric. A dataset comprising 202 samples, including pure turmeric, MY, and their admixtures (5%–40% w/w), was analyzed using both the portable device and a high-resolution benchtop visible and near-infrared–short-wave infrared (VNIR–SWIR) spectroradiometer. Multiple spectral pre-treatments were applied, and classification was performed using Random Forest (RF), XGBoost, and Support Vector Machines (SVM). The best performance was achieved by the RF model on raw reflectance spectra collected with the miniaturized sensor (98% accuracy, $\kappa$ = 0.97). Key diagnostic wavelengths (e.g., 1495, 1640, 1675, 2155, 2475 nm) linked to MY’s chemical structure were pinpointed through feature importance analysis. This work highlights the potential of portable, low-cost NIR sensors as tools for broadening access to food authentication, enabling rapid, non-destructive testing with a practical limit of detection of 5% w/w for MY adulteration.