EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

HCT-Det: A High-Accuracy End-to-End Model for Steel Defect Detection Based on Hierarchical CNN–Transformer Features

Xiyin Chen iD, Xiaohu Zhang, Yonghua Shi iD, Junjie Pang

Sensors · 21.02.2025 · 25 (5) · 1333

DOI: 10.3390/s25051333 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 21.02.2025. Verfasst von Xiyin Chen und 3 weiteren Autor:innen. Beteiligte Einrichtung: School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou 510640, China. Die Arbeit wurde bisher 24-mal zitiert (Crossref) und nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Key R&D Program of China, Guangxi Key R&D Program, Shenzhen Longhua District 2023 Special Fund for Scientific and Technological Innova-tion.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Surface defect detection is essential for ensuring the quality and safety of steel products. While Transformer-based methods have achieved state-of-the-art performance, they face several limitations, including high computational costs due to the quadratic complexity of the attention mechanism, inadequate detection accuracy for small-scale defects due to substantial downsampling, inconsistencies between classification scores and localization confidence, and feature resolution loss caused by simple upsampling and downsampling strategies. To address these challenges, we propose the HCT-Det model, which incorporates a window-based self-attention residual (WSA-R) block structure. This structure combines window-based self-attention (WSA) blocks to reduce computational overhead and parallel residual convolutional (Res) blocks to enhance local feature continuity. The model’s backbone generates three cross-scale features as encoder inputs, which undergo Intra-Scale Feature Interaction (ISFI) and Cross-Scale Feature Interaction (CSFI) to improve detection accuracy for targets of various sizes. A Soft IoU-Aware mechanism ensures alignment between classification scores and intersection-over-union (IoU) metrics during training. Additionally, Hybrid Downsampling (HDownsample) and Hybrid Upsampling (HUpsample) modules minimize feature degradation. Our experiments demonstrate that HCT-Det achieved a mean average precision (mAP@0.5) of 0.795 on the NEU-DET dataset and 0.733 on the GC10-DET dataset, outperforming other state-of-the-art approaches. These results highlight the model’s effectiveness in improving computational efficiency and detection accuracy for steel surface defect detection.

Schlagwörter

Stichwörter der Autor:innen

convolutional neural networkdefect detectiontransformer

Autor:innen und Institutionen

  • School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou 510640, China.
  • Hymson Laser Technology Group Co., Ltd., Shenzhen 518110, China.

ORCID-Profile: Xiyin Chen, Yonghua Shi

Förderung

  • National Key R&D Program of China
  • Guangxi Key R&D Program
  • Shenzhen Longhua District 2023 Special Fund for Scientific and Technological Innova-tion
  • Shenzhen Longhua District 2023 Special Fund for Scientific and Technological Innovation
  • National Key R&D Program of China
  • Guangxi Key R&D Program

Zitieren

APA

Chen, X., Zhang, X., Shi, Y., & Pang, J. (2025). HCT-Det: A High-Accuracy End-to-End Model for Steel Defect Detection Based on Hierarchical CNN–Transformer Features. Sensors, 25(5), 1333. https://doi.org/10.3390/s25051333

Vancouver

Chen X, Zhang X, Shi Y, Pang J. HCT-Det: A High-Accuracy End-to-End Model for Steel Defect Detection Based on Hierarchical CNN–Transformer Features. Sensors. 2025;25(5):1333. doi:10.3390/s25051333

BibTeX
@article{Chen2025,
  author = {Chen, Xiyin and Zhang, Xiaohu and Shi, Yonghua and Pang, Junjie},
  title = {HCT-Det: A High-Accuracy End-to-End Model for Steel Defect Detection Based on Hierarchical CNN–Transformer Features},
  journal = {Sensors},
  year = {2025},
  volume = {25},
  number = {5},
  pages = {1333},
  doi = {10.3390/s25051333}
}
Kostenlos starten →