This work addresses the growing civilian applications of drone technology and the associated security risks, particularly with respect to public and air safety. To address these issues, we introduce an improved YOLOv11-DEC model for real-time detection of small-target unmanned aerial vehicles (UAVs) in complex environments. The model incorporates SBA, group normalisation, and detail-enhanced convolution, and integrates RepViT with EMA attention to increase feature representation and detection accuracy. Additionally, we developed a lightweight LSDECD detection head to maintain model size while preserving accuracy. Experiments on the DUT dataset show that YOLOv11-DEC outperforms baseline models such as YOLOv11, in terms of detection accuracy, offering an effective solution for UAV monitoring and defence systems.
YOLOv11-DEC: An Improved YOLOv11 Model for UAV Detection in Complex Contexts
05.05.2025
1506509 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Traffic accident detection based on YOLOv11
IEEE | 2024
|A Lightweight Anti-Unmanned Aerial Vehicle Detection Method Based on Improved YOLOv11
DOAJ | 2024
|Autonomous UAV Detection of Ochotona curzoniae Burrows with Enhanced YOLOv11
DOAJ | 2025
|ArXiv | 2024
|