Scholarlink

<p>This study builds on our earlier work applying first-generation Gamma Computed Tomography (fgen-GCT) to reinforced concrete. We systematically evaluate six classical edge detectors on super-resolved GCT images and find that a Sobel-based workflow—augmented by grayscale inversion, Otsu thresholding, and morphological closing—provides the most reliable boundaries for sizing reinforcement. Across two phantom images, dimensional errors relative to nominal bar sizes were reduced to 0.5–2.8%, improving on our prior SRGAN-only results (2–5%). The approach is lightweight (~ 0.1&#xa0;s per slice on a CPU), reproducible, and complements AI-based enhancement, advancing fgen-GCT toward practical structural integrity assessment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Edge detection for dimensional characterization in AI-enhanced gamma tomography of reinforced concrete

  • 期刊论文

  • 发表日期：

    2025年9月26日

查看全文
  • Wilson Macharia Kairu,
  • Rabie Outayad,
  • Siphila Wanjiku Mumenya,
  • Kenneth Njoroge,
  • Prabhu Rajagopal

摘要

This study builds on our earlier work applying first-generation Gamma Computed Tomography (fgen-GCT) to reinforced concrete. We systematically evaluate six classical edge detectors on super-resolved GCT images and find that a Sobel-based workflow—augmented by grayscale inversion, Otsu thresholding, and morphological closing—provides the most reliable boundaries for sizing reinforcement. Across two phantom images, dimensional errors relative to nominal bar sizes were reduced to 0.5–2.8%, improving on our prior SRGAN-only results (2–5%). The approach is lightweight (~ 0.1 s per slice on a CPU), reproducible, and complements AI-based enhancement, advancing fgen-GCT toward practical structural integrity assessment.