With the rapid growth of UAVs, ensuring safe and compliant operations is critical. Identifying drones accurately and distinguishing authorized from unauthorized flights presents a major challenge for Unmanned Traffic Management (UTM).
UAVMS integrates advanced visual identification with indoor positioning to verify drone identities.
Visually verifies drone's reported location by fusing AI vision with IPS data. Our AI model detects the drone via camera, then UAVMS compares these "seen" visual coordinates against the "claimed" IPS coordinates.
Powered by YOLOv8 and trained on our custom dataset of 50,000+ images, achieving 90.5% mAP for robust drone detection.
Utilizing Marvelmind IPS, we achieve ±8cm static accuracy, simulating GPS/RID data for indoor flight and verification.
Our core innovation: UAVMS intelligently compares visual data with reported positions to confirm drone identity and flag discrepancies.
mAP Visual Detection Accuracy
Static IPS Accuracy
Frame Custom Training Dataset
Successful Visual-Positional Data Fusion
We see a generally low linear relationship between the XYZ coordinates reported by the IPS and the VID system during dynamic testing
The X-axis deviation between drone's visual position and its IPS-reported position shows moderate accuracy with notable outliers (red zone) during rapid movement phases.
Y-axis comparison reveals similar patterns to X-axis, with acceptable correlation during steady flight but increased discrepancies when the drone changes direction or speed.
Visual Systems & Integration Lead
AI & Computer Vision Specialist
Positioning Systems Engineer
Data Fusion & Algorithm Developer
Professor of Computer Engineering
Research Assistant
Research Assistant
Research Advisor
Questions? Reach out:
100061340@ku.ac.ae