A Real-Time Edge Computing System for Monitoring Bees at Flowers

Honey bees are vital to global agriculture, playing a key role in pollinating crops and supporting our food supply. Understanding their foraging behaviors has traditionally involved labor-intensive field experiments, which often come with the risk of human error, especially when tracking and identifying individual bees within large groups. This paper introduces a system that uses artificial intelligence (AI) models, along with an NVIDIA Jetson Xavier for edge computing, to enhance real-time detection, tracking, and data processing of honey bee flower visits in the field. Our system combines AI detection and a specially designed video processing pipeline, offering a practical solution with a user-friendly interface for field biologists. This approach not only makes field experiments more efficient and feasible, but also enables precise, real-time video data processing, crucial for making on-the-spot decisions during experiments. This article delves into the methodology, performance, and potential future work of the system, showcasing how this combination of technology and biology opens new possibilities for conducting accurate and high-throughput field experiments, ultimately improving our understanding and management of honey bee populations.