Artificial intelligence-based pollen classification machine in apiculture: design, implementation and evaluation

Area of Science:

• Agricultural Science
• Computer Science
• Food Science

Background:

• Bee pollen is a valuable bioactive substance with nutritional and health benefits.
• Variability in botanical origin affects bee pollen's composition, hindering quality control and classification.
• Artificial intelligence (AI) presents a solution for automated and standardized classification of bee pollen.

Purpose of the Study:

• To develop and evaluate an AI-driven system for classifying bee pollen based on color properties.
• To assess the performance of various deep learning models for pollen classification.
• To validate the system's reliability for differentiating monofloral pollen types.

Main Methods:

• A deep learning system was designed to classify pollen samples using color as a biomarker.
• Convolutional neural networks (CNNs) including MobileNet, InceptionV3, Xception, NasNet Large, DenseNet201, and YOLOv8 were evaluated.
• Laboratory-scale validation was performed to confirm the system's accuracy.

Main Results:

• DenseNet201 achieved the highest classification accuracy at 98.5%.
• YOLOv8 demonstrated real-time performance with 91.4% accuracy and rapid processing.
• The system reliably differentiated between various monofloral pollen types.

Conclusions:

• The AI-based classification system offers a robust method for standardizing and tracing pollen products.
• Real-time classification provides beekeepers with a practical tool for sustainable and hygienic pollen collection.
• Potential applications exist in the food, pharmaceutical, nutraceutical, and cosmetic industries.