Varroa Mite Counting Based on Hyperspectral Imaging

Area of Science:

• Agricultural Science
• Computer Science
• Entomology

Background:

• Varroa mite infestations are a critical global threat to honeybee colony survival.
• Accurate Varroa mite quantification is essential for effective pest management strategies.

Purpose of the Study:

• To evaluate the feasibility of using HS-Cam and machine learning for automated Varroa mite counting.
• To compare the performance of hyperspectral imaging against traditional RGB imaging for Varroa detection.

Main Methods:

• Image acquisition using HS-Cam, followed by Principal Component Analysis (PCA) for dimensionality reduction.
• Application of k-Nearest Neighbors (kNNs) for object segmentation and Support Vector Machine (SVM) for shape detection.
• Development of a counting algorithm integrating SVM outputs for quantifying Varroa mites in hyperspectral images.

Main Results:

• Achieved segmentation accuracy exceeding 99% with high precision (0.9983) and recall (0.9947).
• Demonstrated strong agreement between automated counts and manual ground truth.
• HS-Cam showed superior performance for Varroa counting compared to RGB images, even with a limited dataset.

Conclusions:

• The HS-Cam and machine learning approach is a feasible and highly accurate method for Varroa mite counting.
• Hyperspectral imaging presents a promising alternative to traditional methods for monitoring honeybee health.
• Future work includes dataset expansion, exploring near-infrared (NIR) excitation, and smartphone integration.