Detecting Drones at an Entrance to a Beehive Based on Audio Signals and Autoencoder Neural Networks

Detecting a large number of drones at the entrance of a beehive in the late Spring, may be one of the signs of swarming, since the peak of a drone number is correlated with the overcrowded hive. Swarming is a natural method of reproduction observed in honey bees when half of the bee colony leaves their hive with an old queen. This is a potentially dangerous situation, as the queen and the part of the colony may escape. To prevent this, we propose an early swarming detection system based on the behavior of drones near the hive entrance. The system analyzes sound signals recorded at the entrance of the hive by an autoencoder neural network to detect drones. Simulations using real signals have shown that it is possible to effectively detect drones based on audio signals only. The achieved detection accuracy makes it possible to create an effective alarm system for beekeepers.

I. Introduction

Swarming is a natural method of reproduction observed in honey bees [1], [2]. Worker bees are able to detect when it is time to swarm due to overcrowding of the hive or the lack of pheromone production from the queen. A swarm typically contains about half the workers together with the old queen, while the new queen stays back with the remaining workers in the original hive.