An Internet-of-Drones Architecture for Context-Aware Aerial Surveillance

Bernardo De Melo Coimbra Amaral Albano

Key information

Authors:

Bernardo De Melo Coimbra Amaral Albano (Bernardo De Melo Coimbra Amaral Albano)

Supervisors:

António Manuel Raminhos Cordeiro Grilo (António Manuel Raminhos Cordeiro Grilo)

Published in

July 11, 2025

Abstract

This thesis presents the design and implementation of a UAV-based surveillance system that enhances vision-based person detection through the integration of cooperative geolocation IP beacons. The system was developed under the IST Capstone framework and targets real-world scenarios requiring remote situational awareness, such as industrial site monitoring and maritime security. The proposed architecture combines a hybrid VTOL quadplane UAV with a gimballed RGB–thermal camera, onboard GPU-based inference, and LTE/NR-based communication. To move beyond class-level object detection, the system introduces lightweight and low-cost geolocation beacons using LTE-M/NB-IoT and mutual TLS to transmit GNSS data securely to a centralized server. These positions are fused with UAV pose, gimbal orientation, and camera zoom level to project beacon locations into the camera frame in real time. Detected individuals are classified as “authorized” or “intruder” based on proximity to known beacons in the pixel domain, using a threshold informed by GNSS error modeling. A complete IP-based control framework was implemented, including a reverse SSH tunnel, WebSocket-based gimbal command proxy, and a multi-client, adaptive-bitrate video stream using GStreamer and WebRTC. Experimental ground tests validated the system’s classification accuracy, GNSS modeling, and low-latency communication performance. Results demonstrate that the system can correctly update target classification as beacon positions change, even under update delays and moderate GNSS noise. This work provides a practical, extensible foundation for deploying context-aware UAV surveillance solutions in outdoor environments without relying on fixed infrastructure.