EUVIMEDEUVIMED entdecken
Crossref

SFE-fine: Scale-Decoupled Feature Enhancement for wafer map defect detection

Xu Shi, Haijun Liu

Measurement Science and Technology · 17.07.2026 · 37 (28) · 286303

DOI: 10.1088/1361-6501/ae856b ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 14.07.2026. Verfasst von Xu Shi und 1 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Gefördert von: Scientific Research and Development Fund of Hefei University of Technology.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Wafer surface defect detection is an essential part of the semiconductor manufacturing process for ensuring yield and quality. However, as the integration density of wafers increases, wafer surface defect detection faces challenges of missed detection of tiny defects due to low signal-to-noise ratios as well as the geometric truncation of elongated defects due to the limited receptive fields of traditional square convolutions. To overcome these challenges, an improved DEIM framework with Scale-Decoupled Feature Enhancement for wafer surface defect detection is proposed. First, we designed a Micro-Scale Enhancement Module incorporating the SA-PKI structure in shallow layers, where the CSAG gating unit could effectively suppress background noise via a direct gating strategy, and the PKI unit could reconstruct fine-grained texture features for small targets. Second, we introduced a parallel S-LSKA module in deep layers, combining large selective kernel attention with strip pooling. This design implements geometric decoupling, where LSKA captures isotropic global context and strip pooling establishes anisotropic long-range dependencies, particularly improving the detection of scratches with extreme aspect ratios. Experimental results indicate that our method achieves an mAP50 of 98.3 and an mAP50:95 of 98.1 on the wafer surface defect detection task, outperforming the baseline network by 0.8 and 1.2, respectively.

Förderung

  • Scientific Research and Development Fund of Hefei University of Technology

Zitieren

APA

Shi, X., & Liu, H. (2026). SFE-fine: Scale-Decoupled Feature Enhancement for wafer map defect detection. Measurement Science and Technology, 37(28), 286303. https://doi.org/10.1088/1361-6501/ae856b

Vancouver

Shi X, Liu H. SFE-fine: Scale-Decoupled Feature Enhancement for wafer map defect detection. Measurement Science and Technology. 2026;37(28):286303. doi:10.1088/1361-6501/ae856b

BibTeX
@article{Shi2026,
  author = {Shi, Xu and Liu, Haijun},
  title = {SFE-fine: Scale-Decoupled Feature Enhancement for wafer map defect detection},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {28},
  pages = {286303},
  doi = {10.1088/1361-6501/ae856b}
}
Kostenlos starten →