Thermal image processing: A novel approach to predicting honey quality

Abstract

Abstract This study aimed to estimate the proline and Brix content of honey based on the color properties. In the study, 20 honey samples from Türkiye were used, the proline and Brix content of which were determined using the IHC method and the refractometer. The color properties were determined using a spectrophotometer and the ImageJ program, which processed thermal images of honey. Brix and proline estimations were made using the classification and regression trees (CART) algorithm by taking parent node 6, child node 3 and cross validation 5. In the study, the mean Brix and proline values of honeys were 83.2 ± 0.79% and 678.83 ± 192.16 mg/kg, respectively. Analysis of the T-test revealed that there were statistical differences between high and low Brix honeys in terms of L and b values as measured by the spectrophotometer (p < 0.05). The dark and bluish honeys had a lower Brix value and therefore lower sugar content than light and yellowish honeys. In the CART, sensitivity, specificity, and accuracy were 100.00, 0.9091 and 0.9500 for proline and 100.00, 100.00 and 100.00 for Brix, respectively (p < 0.01). The L and b* values of high-proline honeys were higher and lower than 48.143 and 35.416, respectively. Low-Brix honeys had an L and a* value below 55.860 and 53.660 respectively and freshly harvested. As a result, high-proline honeys are whitish and bluish, and low-Brix honeys are more blackish and greenish. However, in order to determine the quality of honey based on color parameters, more extensive studies on honeys from different origins are required.