Towards Varroa destructor mite detection using a narrow spectra illumination

TL;DR

The study addresses Varroa destructor detection on bees entering beehives by leveraging narrow-spectrum, multispectral illumination and a modified beehive monitoring device. It collects a dataset of dead bees and mites under three illumination conditions and trains a U-net semantic segmentation model on infrared data to distinguish mites on bees, comparing it against a conventional computer vision baseline. The approach achieves a pixel-level mite detection probability of about 0.55 ($P \approx 0.55$), with a demonstrated trade-off between false positives and false negatives, and highlights practical deployment challenges such as inference speed on commodity hardware. The work suggests that integrating additional illumination bands and refined losses could improve accuracy, and it outlines steps toward real-world application including a live-bee dataset and hardware acceleration for faster processing.

Abstract

This paper focuses on the development and modification of a beehive monitoring device and Varroa destructor detection on the bees with the help of hyperspectral imagery while utilizing a U-net, semantic segmentation architecture, and conventional computer vision methods. The main objectives were to collect a dataset of bees and mites, and propose the computer vision model which can achieve the detection between bees and mites.

Figures (6)

• Figure 1: Fčielka-Thor 3000 mounted on the beehive.
• Figure 2: LED illumination unit.
• Figure 3: Measured LED spectra.
• Figure 4: Example of photos in dataset.
• Figure 5: Output of the U-net model.