As part of the AI4CUAV project’s mission to enhance UAV detection and behavior recognition, a major milestone was reached with the successful execution of a structured drone recognition trial in Moldova. The field trial, conducted on July 9th, 2025, was designed to evaluate the AI framework’s ability to interpret drone behavior based on elevation, distance, and camera orientation.
The testing scenario involved:
- Predefined flight elevations (5 m, 10 m, 15 m, 20 m)
- A 120-meter test path segmented into four measurable zones
- Multiple drone types including DJI Matrice 300RTK, Matrice 30T, and Mavic 3
- Multi-camera recording setup (frontal and lateral views)
- A custom-built synchronization light system to ensure frame-accurate alignment
This trial forms the core of our scenario-based dataset, which is now being used to benchmark recognition performance of the AI4CUAV models, particularly under real-world environmental and lighting conditions.
Despite early challenges during a failed trial attempt in March 2025 (due to unsynchronized video feeds), the July trial introduced improved synchronization techniques and demonstrated the feasibility of creating measurable and repeatable evaluation environments for behavior-aware drone detection.
📊 What’s Next?
A follow-up trial is planned for Rome in October 2025, where speed variations will be systematically introduced to further analyze dynamic UAV behavior and enhance the robustness of the AI framework.
Stay tuned as we continue pushing the boundaries of trustworthy, interpretable UAV threat detection!
