Related Experiment Videos
MLE-YOLOv11n: Multi-scale layer aggregation and context-aware feature fusion for insulator defect detection in aerial
Yuan Wang1, Hanzhi Cui1, Jinxian Li1
1School of Computing and Intelligent Technology, Qingdao City University, Qingdao, China.
Plos One
|August 3, 2026
Summary
This study introduces MLE-YOLOv11n, an advanced AI model for detecting defects in power transmission insulators using drones. It significantly improves accuracy in identifying complex defects while maintaining high-speed performance for efficient power grid inspection.
Area of Science:
- Electrical Engineering
- Computer Vision
- Artificial Intelligence
Background:
- Power transmission insulators degrade due to high-voltage stress, thermal cycling, and environmental factors, leading to failures.
- Unmanned aerial vehicle (UAV)-based inspection faces challenges with small defect detection, background clutter, and multi-scale features.
Purpose of the Study:
- To develop an enhanced object detection framework for UAV-based insulator defect detection.
- To address limitations in scale, clutter, and feature misalignment in existing lightweight architectures.
Main Methods:
- Proposed MLE-YOLOv11n (Multi-scale Layer aggregation and context-aware feature fusion Enhanced YOLOv11n) framework.
- Integrated SPPELAN, MFCA attention, and Mamba-based MLLA blocks into the YOLOv11n architecture.
- Utilized two public datasets (CPLID and IDID) for evaluation.
Main Results:
- MLE-YOLOv11n achieved 92.1% mAP@50 on CPLID and 95.3% mAP@50 on IDID, outperforming baseline YOLOv11n by 4.4% and 5.2%.
- Significant improvements were observed for rare and complex defect categories.
- The model maintains 85 FPS inference speed with a 10.8% increase in parameters.
Conclusions:
- MLE-YOLOv11n offers a competitive accuracy-efficiency trade-off for UAV-based power grid inspection.
- The framework's performance is suitable for resource-constrained edge platforms.
- The enhanced model effectively detects challenging insulator defects, improving inspection reliability.