EUVIMEDEUVIMED entdecken
Crossref

MSFBHNet: a multiscale spatial-frequency Bayesian-inspired hybrid network for concrete surface defect identification

Chengshang Wu, Tian Zhang, Yiheng Qian, Xiaohan Zhang

Surface Topography: Metrology and Properties · 01.12.2026 · 14 (4) · 045010

DOI: 10.1088/2051-672x/aeba04 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Surface Topography: Metrology and Properties, online erschienen am 04.11.2026. Verfasst von Chengshang Wu und 3 weiteren Autor:innen. Die Arbeit nennt 54 Literaturangaben. Gefördert von: Science and Technology Plan of Liaoning Province in China, Transportation Department of Liaoning Province in China, National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Concrete surface defect identification is crucial for infrastructure safety assessment and subsequent maintenance. However, existing methods still exhibit instability when faced with challenges such as weakened texture and strong noise interference. To address this, we propose MSFBHNet, a novel multiscale spatial-frequency Bayesian-inspired hybrid network designed for robust identification of concrete surface defects. Specifically, we first extract morphological variation information under different receptive fields through the Multiscale Asymmetric Convolution block, enabling the model to effectively represent concrete surface damage features at different scales and morphologies. Subsequently, we design a Global-Directional Frequency Attention module to capture global spectral semantics and directional energy distribution features, and utilize this information to adaptively modulate spatial features, thereby enhancing the model’s robustness under complex backgrounds and noise interference. Furthermore, we introduce a Bayesian-inspired Distributional Prediction module featuring learnable Gaussian classifier parameters. Reparameterization sampling is employed during training, while Monte Carlo inference aggregates predictions from multiple sampled classifiers to obtain more robust predictions. Extensive experiments on three authoritative public datasets indicate that the proposed method achieves superior results in multiple metrics, including recall, F 1 score, and accuracy, while requiring only 3.88 million parameters and 0.46 GMACs. This demonstrates that the proposed method achieves a favorable balance between identification performance and computational efficiency.

Förderung

  • Science and Technology Plan of Liaoning Province in China
  • Transportation Department of Liaoning Province in China
  • National Natural Science Foundation of China
  • Open Project of National Engineering Research Center of High-speed Railway Construction Technology in China

Zitieren

APA

Wu, C., Zhang, T., Qian, Y., & Zhang, X. (2026). MSFBHNet: a multiscale spatial-frequency Bayesian-inspired hybrid network for concrete surface defect identification. Surface Topography: Metrology and Properties, 14(4), 045010. https://doi.org/10.1088/2051-672x/aeba04

Vancouver

Wu C, Zhang T, Qian Y, Zhang X. MSFBHNet: a multiscale spatial-frequency Bayesian-inspired hybrid network for concrete surface defect identification. Surface Topography: Metrology and Properties. 2026;14(4):045010. doi:10.1088/2051-672x/aeba04

BibTeX
@article{Wu2026,
  author = {Wu, Chengshang and Zhang, Tian and Qian, Yiheng and Zhang, Xiaohan},
  title = {MSFBHNet: a multiscale spatial-frequency Bayesian-inspired hybrid network for concrete surface defect identification},
  journal = {Surface Topography: Metrology and Properties},
  year = {2026},
  volume = {14},
  number = {4},
  pages = {045010},
  doi = {10.1088/2051-672x/aeba04}
}
Kostenlos starten →