A Two-Stage YOLOv5s–U-Net Framework for Defect Localization and Segmentation in Overhead Transmission Lines

Transmission-line defect detection is crucial for grid operation. Existing methods struggle to balance defect localization and fine segmentation. Therefore, this study proposes a novel cascaded two-stage framework that first utilizes YOLOv5s for the global localization of defective regions, and then...

Full description

Saved in:
Bibliographic Details
Main Authors: Aohua Li, Dacheng Li, Anjing Wang
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/9/2903
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850031952672849920
author Aohua Li
Dacheng Li
Anjing Wang
author_facet Aohua Li
Dacheng Li
Anjing Wang
author_sort Aohua Li
collection DOAJ
description Transmission-line defect detection is crucial for grid operation. Existing methods struggle to balance defect localization and fine segmentation. Therefore, this study proposes a novel cascaded two-stage framework that first utilizes YOLOv5s for the global localization of defective regions, and then uses U-Net for the fine segmentation of candidate regions. To improve the segmentation performance, U-Net adopts a transfer learning strategy based on the VGG16 pretrained model to alleviate the impact of limited dataset size on the training effect. Meanwhile, a hybrid loss function that combines Dice Loss and Focal Loss is designed to solve the small-target and class imbalance problems. This method integrates target detection and fine segmentation, enhancing detection precision and improving the extraction of detailed damage features. Experiments on the self-constructed dataset show that the method achieves 87% mAP on YOLOv5s, 88% U-Net damage recognition precision, a mean Dice coefficient of 93.66%, and 89% mIoU, demonstrating its effectiveness in accurately detecting transmission-line defects and efficiently segmenting the damage region, providing assistance for the intelligent operation and maintenance of transmission lines.
format Article
id doaj-art-125cedb256784ffbbceefcd4afd98a08
institution DOAJ
issn 1424-8220
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-125cedb256784ffbbceefcd4afd98a082025-08-20T02:58:48ZengMDPI AGSensors1424-82202025-05-01259290310.3390/s25092903A Two-Stage YOLOv5s–U-Net Framework for Defect Localization and Segmentation in Overhead Transmission LinesAohua Li0Dacheng Li1Anjing Wang2School of Mechanical Engineering, Hefei University of Technology, Hefei 230009, ChinaAnhui Institute of Optics and Fine Mechanics, Key Laboratory of General Optical Calibration and Characterization Technology, Chinese Academy of Sciences, Hefei 230031, ChinaAnhui Institute of Optics and Fine Mechanics, Key Laboratory of General Optical Calibration and Characterization Technology, Chinese Academy of Sciences, Hefei 230031, ChinaTransmission-line defect detection is crucial for grid operation. Existing methods struggle to balance defect localization and fine segmentation. Therefore, this study proposes a novel cascaded two-stage framework that first utilizes YOLOv5s for the global localization of defective regions, and then uses U-Net for the fine segmentation of candidate regions. To improve the segmentation performance, U-Net adopts a transfer learning strategy based on the VGG16 pretrained model to alleviate the impact of limited dataset size on the training effect. Meanwhile, a hybrid loss function that combines Dice Loss and Focal Loss is designed to solve the small-target and class imbalance problems. This method integrates target detection and fine segmentation, enhancing detection precision and improving the extraction of detailed damage features. Experiments on the self-constructed dataset show that the method achieves 87% mAP on YOLOv5s, 88% U-Net damage recognition precision, a mean Dice coefficient of 93.66%, and 89% mIoU, demonstrating its effectiveness in accurately detecting transmission-line defects and efficiently segmenting the damage region, providing assistance for the intelligent operation and maintenance of transmission lines.https://www.mdpi.com/1424-8220/25/9/2903transmission-line defectsdefect localizationsemantic segmentationYOLOv5sU-nettwo-stage framework
spellingShingle Aohua Li
Dacheng Li
Anjing Wang
A Two-Stage YOLOv5s–U-Net Framework for Defect Localization and Segmentation in Overhead Transmission Lines
Sensors
transmission-line defects
defect localization
semantic segmentation
YOLOv5s
U-net
two-stage framework
title A Two-Stage YOLOv5s–U-Net Framework for Defect Localization and Segmentation in Overhead Transmission Lines
title_full A Two-Stage YOLOv5s–U-Net Framework for Defect Localization and Segmentation in Overhead Transmission Lines
title_fullStr A Two-Stage YOLOv5s–U-Net Framework for Defect Localization and Segmentation in Overhead Transmission Lines
title_full_unstemmed A Two-Stage YOLOv5s–U-Net Framework for Defect Localization and Segmentation in Overhead Transmission Lines
title_short A Two-Stage YOLOv5s–U-Net Framework for Defect Localization and Segmentation in Overhead Transmission Lines
title_sort two stage yolov5s u net framework for defect localization and segmentation in overhead transmission lines
topic transmission-line defects
defect localization
semantic segmentation
YOLOv5s
U-net
two-stage framework
url https://www.mdpi.com/1424-8220/25/9/2903
work_keys_str_mv AT aohuali atwostageyolov5sunetframeworkfordefectlocalizationandsegmentationinoverheadtransmissionlines
AT dachengli atwostageyolov5sunetframeworkfordefectlocalizationandsegmentationinoverheadtransmissionlines
AT anjingwang atwostageyolov5sunetframeworkfordefectlocalizationandsegmentationinoverheadtransmissionlines
AT aohuali twostageyolov5sunetframeworkfordefectlocalizationandsegmentationinoverheadtransmissionlines
AT dachengli twostageyolov5sunetframeworkfordefectlocalizationandsegmentationinoverheadtransmissionlines
AT anjingwang twostageyolov5sunetframeworkfordefectlocalizationandsegmentationinoverheadtransmissionlines