Leveraging RegNet and CBAM for precise detection of honey adulteration using thermal image analysis

Area of Science:

• Food Science
• Analytical Chemistry
• Computer Science

Background:

• Honey adulteration presents significant health and economic risks.
• Existing detection methods are often slow, expensive, and lack sensitivity.
• There is a need for rapid, accurate, and cost-effective honey quality control.

Purpose of the Study:

• To develop a novel method for classifying honey adulteration levels.
• To utilize thermal imaging and Artificial Intelligence (AI) for honey authenticity verification.
• To overcome limitations of traditional honey quality assessment techniques.

Main Methods:

• A dataset of thermal images was created for pure and adulterated honey samples (1-30% glucose syrup).
• An adaptable Artificial Intelligence (AI) model was developed for classifying adulteration.
• The model's accuracy, sensitivity, and specificity were evaluated across different adulteration levels.

Main Results:

• The AI model achieved 100% precision and specificity for pure honey and 1% adulteration.
• High performance was observed at higher adulteration levels (0.98 for 3%, 0.97 for 5%).
• The method demonstrated swift identification capabilities and versatility across honey varieties.

Conclusions:

• Thermal imaging combined with AI offers a reliable solution for honey quality control.
• This approach enhances the verification of natural bee product authenticity and safety.
• The methodology improves quality assurance, boosting consumer confidence in honey products.