Cygnus: A Vision-Based Drone System for Drowning Detection Using IoT

This paper presents Cygnus, an innovative UAV-based system designed to detect drowning incidents in real time using vision-based techniques. By fusing RGB and infrared imagery and deploying a YOLOv8 model on GAP8 edge microcontrollers onboard drones, Cygnus enables robust monitoring even in low-light or complex aquatic environments.

The system integrates a LoRa-based wireless sensor network (WSN) for drone-to-drone and drone-to-server communication, an on-board image processing pipeline, and a lifeguard-facing web dashboard. A DroneZoom technique is employed to optimize image resolution in ambiguous situations by adjusting drone altitude dynamically.

Cygnus achieves a precision of 95% and a mean Average Precision (mAP) of 83% on a custom dataset of 1000 annotated aquatic activity images. Its modular architecture includes sensor fusion, onboard edge processing, autonomous drone routing, and LoRa WAN coordination — making it a powerful tool for enhancing water safety and enabling rapid response during emergencies.

Authors: Dhumravarna Ambre; Hariharan Sureshkumar*; Prasiddh Trivedi
Download Paper