Unmasking honey adulteration : a breakthrough in quality assurance through cutting-edge convolutional neural network analysis of thermal images

Honey, a natural product generated from organic sources, is widely recognized for its revered reputation. Nevertheless, honey is susceptible to adulteration, a situation that has substantial consequences for both the well-being of the general population and the financial well-being of a country. Conventional approaches for detecting honey adulteration are often associated with extensive time requirements and restricted sensitivity. This paper presents a novel approach to address the aforementioned issue by employing Convolutional Neural Networks (CNNs) for the classification of honey samples based on thermal images. The use of thermal imaging technique offers a significant advantage in detecting adulterants, as it can reveal differences in temperature in honey samples caused by variations in sugar composition, moisture levels, and other substances used for adulteration. To establish a meticulous approach to categorizing honey, a thorough dataset comprising thermal images of authentic and tainted honey samples was collected. Several state-of-the-art Convolutional Neural Network (CNN) models were trained and optimized using the dataset that was gathered. Within this set of models, there exist pre-trained models such as InceptionV3, Xception, VGG19, and ResNet that have exhibited exceptional performance, achieving classification accuracies ranging from 88% to 98%. Furthermore, we have implemented a more streamlined and less complex convolutional neural network (CNN) model, outperforming comparable models with an outstanding accuracy rate of 99%. This simplification offers not only the sole advantage of the model, but it also concurrently offers a more efficient solution in terms of resources and time. This approach offers a viable way to implement quality control measures in the honey business, so guaranteeing the genuineness and safety of this valuable organic commodity.