EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

GM-DETR: Research on a Defect Detection Method Based on Improved DETR

Xin Liu iD, Xudong Yang, Lianhe Shao, Xihan Wang, Quanli Gao iD, Hongbo Shi

Sensors · 03.06.2024 · 24 (11) · 3610

DOI: 10.3390/s24113610 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 03.06.2024. Verfasst von Xin Liu und 5 weiteren Autor:innen. Beteiligte Einrichtung: School of Computer Science, State and Local Joint Engineering Research Center for Advanced Networking & Intelligent Information Services, Xi'an Polyt…. Die Arbeit wurde bisher 27-mal zitiert (Crossref) und nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Natural Science Foundation of China, the Shaanxi Provincial Key Industry Innovation Chain Program, Natural Science Basis Research Plan in Shaanxi Province of China.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Defect detection is an indispensable part of the industrial intelligence process. The introduction of the DETR model marked the successful application of a transformer for defect detection, achieving true end-to-end detection. However, due to the complexity of defective backgrounds, low resolutions can lead to a lack of image detail control and slow convergence of the DETR model. To address these issues, we proposed a defect detection method based on an improved DETR model, called the GM-DETR. We optimized the DETR model by integrating GAM global attention with CNN feature extraction and matching features. This optimization process reduces the defect information diffusion and enhances the global feature interaction, improving the neural network’s performance and ability to recognize target defects in complex backgrounds. Next, to filter out unnecessary model parameters, we proposed a layer pruning strategy to optimize the decoding layer, thereby reducing the model’s parameter count. In addition, to address the issue of poor sensitivity of the original loss function to small differences in defect targets, we replaced the L1 loss in the original loss function with MSE loss to accelerate the network’s convergence speed and improve the model’s recognition accuracy. We conducted experiments on a dataset of road pothole defects to further validate the effectiveness of the GM-DETR model. The results demonstrate that the improved model exhibits better performance, with an increase in average precision of 4.9% (mAP@0.5), while reducing the parameter count by 12.9%.

Schlagwörter

Stichwörter der Autor:innen

DETRGAMdefect detectiontransformer

Autor:innen und Institutionen

  • School of Computer Science, State and Local Joint Engineering Research Center for Advanced Networking & Intelligent Information Services, Xi'an Polytechnic University, Xi'an 710048, China.
  • Shaanxi Province Institute of Water Resources and Electric Power Investigation and Design, Xi'an 710001, China.

ORCID-Profile: Xin Liu, Quanli Gao

Förderung

  • the Natural Science Foundation of China
  • the Shaanxi Provincial Key Industry Innovation Chain Program
  • Natural Science Basis Research Plan in Shaanxi Province of China
  • Xi'an Major Scientific and Technological Achievements Transformation Indus-trialization Project
  • Natural Science Foundation of China
  • Shaanxi Provincial Key Industry Innovation Chain Program
  • Xi’an Major Scientific and Technological Achievements Transformation Industrialization Project

Zitieren

APA

Liu, X., Yang, X., Shao, L., Wang, X., Gao, Q., & Shi, H. (2024). GM-DETR: Research on a Defect Detection Method Based on Improved DETR. Sensors, 24(11), 3610. https://doi.org/10.3390/s24113610

Vancouver

Liu X, Yang X, Shao L, Wang X, Gao Q, Shi H. GM-DETR: Research on a Defect Detection Method Based on Improved DETR. Sensors. 2024;24(11):3610. doi:10.3390/s24113610

BibTeX
@article{Liu2024,
  author = {Liu, Xin and Yang, Xudong and Shao, Lianhe and Wang, Xihan and Gao, Quanli and Shi, Hongbo},
  title = {GM-DETR: Research on a Defect Detection Method Based on Improved DETR},
  journal = {Sensors},
  year = {2024},
  volume = {24},
  number = {11},
  pages = {3610},
  doi = {10.3390/s24113610}
}
Kostenlos starten →