A Transmittance System for Detecting Adulterants in Bee Honey Using Hyperspectral Imaging

Juarez L.; Avila-George H.; Tirado-Kulieva V.A.; Castro W. · SN Computer Science (2026) · DOI: 10.1007/s42979-026-05381-y

Resumen

The adulteration of honey with low-cost syrups compromises its authenticity and hinders quality control because of the physicochemical similarity between the original matrix and the adulterants. This study aimed to quantify honey adulteration with corn syrup and invert sugar using hyperspectral transmittance imaging in the visible and near-infrared (Vis-NIR) range combined with machine learning. A total of 432 experimental preparations derived from 36 original multifloral honey samples were evaluated across two adulteration scenarios, with adulterant concentrations ranging from 0 to 50% (w/w), and spectra were acquired over 399–1002.5 nm. Raw spectra and spectra preprocessed using Standard Normal Variate and Savitzky–Golay filtering (SNV+SG) were modeled using support vector machine regression (SVMR) and artificial neural networks (ANN). Furthermore, -coefficients and the least absolute shrinkage and selection operator (LASSO) reduced the spectral input from 112 to 25 wavelengths. The ANN model with SNV+SG achieved the best performance for corn syrup, with,, and. For invert sugar, the corresponding model achieved,, and. The reduced ANN models maintained values of 0.94 and 0.87 for corn syrup and invert sugar, respectively. Overall, combining hyperspectral imaging, spectral preprocessing, and ANN supported adulterant quantification under the evaluated laboratory conditions, while wavelength selection provided parsimonious models that may inform the future investigation of simplified spectral systems. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2026.

Instituciones: Universidad Nacional de Frontera