| Tartalmi leírás: |
Steel-fiber-reinforced concrete (SFRC) relies on the distribution, orientation, and continuity of embedded fibers, yet these characteristics are difficult to assess non-destructively after hardening. X-ray computed tomography (CT) provides volumetric imaging of internal structures, but converting CT volumes into reliable fiber-level descriptions remains challenging due to imaging noise, partial volume effects, and frequent fiber interactions (contact, overlap, and apparent merging). This vision paper outlines a research agenda centered on practical and reproducible annotation of steel fibers in volumetric CT data. We argue for a hybrid workflow in which automated slice-based detection and inter-slice association provide scalable first-pass annotations, while targeted manual inspection supplies contextual corrections and supports trustworthy validation. The paper identifies core technical gaps (robust separation of touching fibers, stable identity tracking across slices, and uncertainty-aware validation), proposes a modular methodology to address them, and describes an evaluation plan grounded in selective manual annotation, consistency checks, and sensitivity analysis. The overarching goal is to enable transparent CT-based SFRC assessment that is feasible for laboratory-sized datasets while pro- ducing fiber-level descriptors suitable for downstream mechanical interpretation and quality assessment.
|
