Computer Vision and AI Algorithms Edge Computation on UAVs (edgeAI4UAV)
The project “Computer Vision and AI Algorithms Edge Computation on UAVs (edgeAI4UAV)” has been selected as one of the five research initiatives funded through the AI4Media – A European Excellence Centre for Media, Society and Democracy. It is implemented by members of the laboratory under the scientific supervision of Professor Dr. Vasileios Chatzi, in collaboration with researchers from the Departments of Management Science & Technology and Computer Science of the Democritus University of Thrace.
Overview
The primary goal of edgeAI4UAV is to develop innovative computer vision and Artificial Intelligence (deep learning) algorithms capable of running directly on-board Unmanned Aerial Vehicles (UAVs) through edge computing techniques. The system focuses on real-time detection and tracking of people and moving objects, enabling autonomous cinematography tasks and event-driven visual documentation.
Technical Approach
The project integrates:
A stereo camera system, providing depth information via lightweight stereo-to-depth algorithms.
An embedded edge processing unit for real-time onboard computation.
A WiFi module allowing the UAV to transmit selected images to a server while in flight.
The computer vision algorithms are designed to detect and track moving targets, while their outputs are utilized by an embedded decision-making module. This module dynamically adjusts UAV navigation in real time, enabling tasks such as:
Following a specific moving person or object (e.g., actor, animal).
Changing viewpoint angles (side, front, etc.).
Approaching or distancing from the target depending on mission requirements.
Through this integrated perception-and-control framework, the UAV can autonomously execute complex cinematographic missions without continuous human intervention.
Impact
The project advances the field of autonomous aerial cinematography by combining deep learning–based perception with classical path-planning and control methods. It enables high-quality visual content acquisition in dynamic environments, while demonstrating the potential of edge AI for real-time decision-making in UAV platforms.
Main Deliverables
Real-time computer vision algorithms for human detection and tracking from an onboard UAV camera.
A decision-support and control system enabling autonomous UAV navigation.
A fully functional UAV prototype with integrated human-tracking capabilities.
Funding
The edgeAI4UAV project has indirectly received funding from the European Union’s Horizon 2020 Research and Innovation Programme through the AI4Media Open Call #1, under the AI4Media project (Grant Agreement No. 951911).