EUVIMEDEUVIMED entdecken
Crossref

Insulator defect detection based on multi-scale feature fusion and local topology reconstruction

Zixi Fan, Wei Guo, Haicheng Qu

Measurement Science and Technology · 03.07.2026 · 37 (27) · 275002

DOI: 10.1088/1361-6501/ae7c19 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 01.07.2026. Verfasst von Zixi Fan und 2 weiteren Autor:innen. Die Arbeit nennt 42 Literaturangaben. Gefördert von: Basic Research Project of the Education Department of Liaoning Province, General Program of the National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Insulator defect detection is crucial for ensuring the normal operation of power transmission and distribution networks. However, challenges such as the irregular shapes of defects and difficulties in detection under adverse weather conditions arise during insulator defect inspection. To tackle these issues, we propose an insulator defect detection approach based on multi-scale feature fusion and local topology reconstruction. Firstly, we designed a CSP-PMS module, which effectively mitigates the gradient vanishing problem in deep networks through staged feature segmentation and recombination, while significantly reducing computational redundancy. Secondly, to better fuse the features extracted by the backbone network, we propose a novel local topology reconstruction feature-focused diffusion pyramid network. By leveraging local topology reconstruction, it enables input-content-driven adaptive adjustment of contextual information, thereby breaking through the limitations of static architectures. Additionally, through kernel space projection, it transforms sparse kernels into dense forms, achieving a balance between lightweight design and information integrity. In addition, addressing the issue of insufficient information interaction among traditional detection heads, which can lead to the loss of partial information, we propose a lightweight detail-enhanced convolutional detection head. Finally, we carried out experiments on both our self-collected dataset and public benchmark datasets. The results show that our approach attains an mAP50 of 92.5% on the self-built dataset. On the publicly available CPLID and Pascal VOC datasets, our method achieved mAP50 of 98% and 73%, respectively, while significantly reducing the number of parameters. These results further validate the effectiveness and practicality of our proposed model.

Förderung

  • Basic Research Project of the Education Department of Liaoning Province
  • General Program of the National Natural Science Foundation of China

Zitieren

APA

Fan, Z., Guo, W., & Qu, H. (2026). Insulator defect detection based on multi-scale feature fusion and local topology reconstruction. Measurement Science and Technology, 37(27), 275002. https://doi.org/10.1088/1361-6501/ae7c19

Vancouver

Fan Z, Guo W, Qu H. Insulator defect detection based on multi-scale feature fusion and local topology reconstruction. Measurement Science and Technology. 2026;37(27):275002. doi:10.1088/1361-6501/ae7c19

BibTeX
@article{Fan2026,
  author = {Fan, Zixi and Guo, Wei and Qu, Haicheng},
  title = {Insulator defect detection based on multi-scale feature fusion and local topology reconstruction},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {27},
  pages = {275002},
  doi = {10.1088/1361-6501/ae7c19}
}
Kostenlos starten →