With the increasing deployment of UAVs, accurate and efficient detection is critical for preventing UAV intrusions and ensuring airspace security. However, existing detection algorithms often struggle with large target size variations and limited computational resources. Lightweight detection algorithms are crucial for real-time, efficient UAV detection. This paper presents "NUAV-YOLO," a lightweight object detection algorithm tailored for UAV detection. To address the challenge of scale variation caused by UAVs' differing flight altitudes and distances, NUAV-YOLO employs multi-scale feature fusion. Building on YOLOv8, the algorithm introduces a self-attention-based segmentation detection head to enhance target localization and classification accuracy. Additionally, knowledge distillation is applied to further optimize YOLOv8's performance. Experimental results demonstrate that NUAV-YOLO outperforms existing methods in both detection accuracy and operational efficiency.
NUAV-YOLO: a lightweight object detection algorithm based on YOLOv8 for UAVs
Fourth International Conference on Electronic Information Engineering and Data Processing (EIEDP 2025) ; 2025 ; Kuala Lumpur, Malaysia
Proc. SPIE ; 13574 ; 135740B
2025-05-09
Conference paper
Electronic Resource
English
G-YOLO: A Lightweight Infrared Aerial Remote Sensing Target Detection Model for UAVs Based on YOLOv8
DOAJ | 2024
|GPF-YOLO: A Real-Time Vehicle Object Detection Algorithm Based on Improved YOLOv8
Springer Verlag | 2025
|