Dehydrated bee pollen is a premium product with a growing market as a food supplement, functional food ingredient, and is also an ingredient in biomedical and healthcare formulations. Pollen from certain species of stingless bees is higher priced than the equivalent from the much more common honey bee, as the former is considered to have enhanced nutritional and health benefits. Pollens from different bee species are visually identical, giving an incentive for potential fraud.
In this paper (purchase required) the authors built a classification model to discriminate pollens from different bee species. The researchers integrated digital image processing and machine learning to classify pollen loads produced by honey bees (Apis mellifera) and pot-pollen from stingless bee species based on their colour patterns. A total of 246 pollen loads and pot-pollen samples from five bee species (Apis mellifera, Melipona marginata, Melipona quadrifasciata quadrifasciata, Scaptotrigona bipunctata, and Tetragona clavipes) were collected, and high-resolution images were captured using a smartphone. Colour parameters extracted from images such as R, B, H, and V were analyzed, and classification models employing CatBoost, XGBoost, Random Forest, and k-Nearest Neighbors (kNN) algorithms were tested.
They reported that CatBoost achieved the highest performance, with accuracies of 100 % in the training phase and 98.9 % in the testing phase. Linear Discriminant Analysis (LDA) further validated the classification by grouping the pollen samples into five distinct clusters corresponding to the bee species studied.
They conclude that combining digital image processing from a smartphone with machine learning offers an effective approach to classifying pollen from different bee species, with promising applications in apiculture to ensure product quality and authenticity.
Photo by Aaron Burden on Unsplash
