RSP-YOLOv11n multi-module optimized algorithm for insulator defect detection in UAV images

The identification of insulator defects on transmission line insulators constitutes a pivotal undertaking in the context of UAV (Unmanned Aerial Vehicle) inspection, a process that is imperative to ensure the reliable functioning of transmission lines. A novel approach is proposed to mitigate missed...

Full description

Saved in:
Bibliographic Details
Published in:Scientific reports Vol. 15; no. 1; pp. 35426 - 19
Main Authors: Zheng, Bin, Angkawisittpan, Niwat, Huang, Lu, Sonasang, Somchat
Format: Journal Article
Language:English
Published: London Nature Publishing Group UK 10.10.2025
Nature Publishing Group
Nature Portfolio
Subjects:
639/166/987
639/705/258
Accuracy
Algorithms
Deep learning
Defects
Efficiency
Humanities and Social Sciences
Image processing
Inspection
Inspections
Insulator defect detection
multidisciplinary
P2 detection head
RSP-YOLOv11n
Science
Science (multidisciplinary)
Transmission lines
UAV inspection
Weather
China
RSP-YOLOv11n
UAV inspection
Insulator defect detection
P2 detection head
ISSN:2045-2322, 2045-2322
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The identification of insulator defects on transmission line insulators constitutes a pivotal undertaking in the context of UAV (Unmanned Aerial Vehicle) inspection, a process that is imperative to ensure the reliable functioning of transmission lines. A novel approach is proposed to mitigate missed detections and enhance detection accuracy in UAV-based insulator defect detection. RSP-YOLOv11n (RCSOSA-SEA-P2 YOLOv11n) is proposed to enhance the detection of insulator defects in UAV-acquired imagery. The conventional C3K2 module in the backbone is replaced with the RCSOSA unit, thereby enabling more effective multi-scale feature extraction and representation learning. Second, by using axial attention and detail enhancement, the SEA attention mechanism is used to improve the ability to detect surface defects on insulators. Finally, by capturing finer features during high-resolution image processing, the addition of a P2 detection head to the network improves the accuracy of small target detection. RSP-YOLOv11n performs better overall than other YOLO series models, according to experimental results on the self-constructed insulator dataset. In contrast to the baseline YOLOv11n model, RSP-YOLOv11n improved precision from 89.9 to 92.3%, recall from 82.6 to 85.9%, F1-score from 86.1 to 89.0%, from mAP@0.5 from 88.7 to 91.2%, and mAP@0.5:0.95 from 58.9 to 61.7%. Furthermore, the proposed RSP-YOLOv11n framework was evaluated on three benchmark insulator datasets—CPLID, IDID, and SFID. Across these datasets, it consistently achieved better detection performance compared to other models in the YOLO family. In addition, RSP-YOLOv11n exhibited competitive advantages over recent state-of-the-art detectors, including DINO and RT-DETR. The experimental results highlight the framework’s strong capability in small object detection, showing notable improvements in accuracy and generalization. These findings suggest that RSP-YOLOv11n holds considerable potential for meeting the practical requirements of insulator defect detection in real-world UAV inspection scenarios.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:2045-2322
2045-2322
DOI:10.1038/s41598-025-19059-7