EUVIMEDEUVIMED entdecken
Crossref

Vision–Language Model Assisted Interpretation in Radiographic Welding Defect Evaluation

Haoyu Wen, Baoxin Zhang, Xuefeng Zhao, Juntao Wu, Jiajia Liu, Wangwang Liu, Ruotong Zhang, Hao Qiu, Xinghua Yu

e-Journal of Nondestructive Testing · 2026-07 · 31 (7)

DOI: 10.58286/32968 ↗

Kurz gefasst

Fachartikel in der Zeitschrift e-Journal of Nondestructive Testing, 2026-07. Verfasst von Haoyu Wen und 8 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This work presents an application of vision–language models (VLMs) to assist radiographic welding defect interpretation in industrial non-destructive testing (NDT). The approach explores how visual understanding and linguistic reasoning can be combined to enhance the interpretability and reliability of radiographic evaluation. The system integrates image enhancement, defect localization, and VLM-based descriptive reasoning to support inspectors in identifying and explaining potential defects. Preliminary applications show that VLM-assisted interpretation can provide intuitive textual explanations and visual cues, improving human–machine collaboration in inspection.

Zitieren

APA

Wen, H., Zhang, B., Zhao, X., Wu, J., Liu, J., Liu, W., Zhang, R., Qiu, H., & Yu, X. (2026). Vision–Language Model Assisted Interpretation in Radiographic Welding Defect Evaluation. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/32968

Vancouver

Wen H, Zhang B, Zhao X, Wu J, Liu J, Liu W, et al. Vision–Language Model Assisted Interpretation in Radiographic Welding Defect Evaluation. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/32968

BibTeX
@article{Wen2026,
  author = {Wen, Haoyu and Zhang, Baoxin and Zhao, Xuefeng and Wu, Juntao and Liu, Jiajia and Liu, Wangwang and Zhang, Ruotong and Qiu, Hao and Yu, Xinghua},
  title = {Vision–Language Model Assisted Interpretation in Radiographic Welding Defect Evaluation},
  journal = {e-Journal of Nondestructive Testing},
  year = {2026},
  volume = {31},
  number = {7},
  doi = {10.58286/32968}
}
Kostenlos starten →