EUVIMEDEUVIMED entdecken
Crossref

Projection-Domain Defect Segmentation for Accelerated Quality Assessment in Industrial Computed Tomography

Tim Schanz, Robin Tenscher-Philipp, Martin Simon

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

DOI: 10.58286/33406 ↗

Kurz gefasst

Fachartikel in der Zeitschrift e-Journal of Nondestructive Testing, 2026-07. Verfasst von Tim Schanz und 2 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Industrial computed tomography (CT) is a key technique for non-destructive testing but commonly relies on computationally expensive volumetric reconstruction prior to defect analysis. This work presents a projection-domain defect segmentation approach that enables early-stage quality pre-evaluation without full reconstruction. Deep learning models are trained on physically consistent synthetic projection data, and small angular sinogram stacks are processed to improve robustness under noisy conditions. Experiments on large-scale simulated datasets demonstrate reliable defect segmentation and counting directly in the projection domain, enabling reduced scan and processing time and supporting accelerated CT-based industrial inspection.

Zitieren

APA

Schanz, T., Tenscher-Philipp, R., & Simon, M. (2026). Projection-Domain Defect Segmentation for Accelerated Quality Assessment in Industrial Computed Tomography. e-Journal of Nondestructive Testing, 31(7). https://doi.org/10.58286/33406

Vancouver

Schanz T, Tenscher-Philipp R, Simon M. Projection-Domain Defect Segmentation for Accelerated Quality Assessment in Industrial Computed Tomography. e-Journal of Nondestructive Testing. 2026;31(7). doi:10.58286/33406

BibTeX
@article{Schanz2026,
  author = {Schanz, Tim and Tenscher-Philipp, Robin and Simon, Martin},
  title = {Projection-Domain Defect Segmentation for Accelerated Quality Assessment in Industrial Computed Tomography},
  journal = {e-Journal of Nondestructive Testing},
  year = {2026},
  volume = {31},
  number = {7},
  doi = {10.58286/33406}
}
Kostenlos starten →