Scholarlink

<p>Muon Scattering Tomography (MST) is a non-invasive method that produces three-dimensional voxel grids of scattering density by Points of Closest Approach (POCA) reconstruction. Previous research [<CitationRef CitationID="CR1">1</CitationRef>] concentrated on CNN-based denoising; nevertheless, operational threat identification necessitates automatic 3D object segmentation and material quantification. We introduce a resilient multi-stage deterministic pipeline for object separation and boundary delineation in noisy POCA reconstructions, which does not necessitate machine learning training. Our pipeline integrates threshold-based segmentation with connected component labeling (CCL), employs watershed segmentation for the delineation of nearby objects, and utilizes morphological opening for boundary refinement. This training-free method effectively isolates individual components within the low-resolution MST environment, providing stable segmentation that facilitates volumetric estimation and density characterization for operational MST systems.</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

Robust object segmentation and quantification in muon tomography using a multi-stage deterministic image processing pipeline

  • 期刊论文

  • 发表日期：

    2026年4月28日

查看全文
  • Prasoon Kumar Vinodkumar,
  • Cagri Ozcinar,
  • Gholamreza Anbarjafari

摘要

Muon Scattering Tomography (MST) is a non-invasive method that produces three-dimensional voxel grids of scattering density by Points of Closest Approach (POCA) reconstruction. Previous research [1] concentrated on CNN-based denoising; nevertheless, operational threat identification necessitates automatic 3D object segmentation and material quantification. We introduce a resilient multi-stage deterministic pipeline for object separation and boundary delineation in noisy POCA reconstructions, which does not necessitate machine learning training. Our pipeline integrates threshold-based segmentation with connected component labeling (CCL), employs watershed segmentation for the delineation of nearby objects, and utilizes morphological opening for boundary refinement. This training-free method effectively isolates individual components within the low-resolution MST environment, providing stable segmentation that facilitates volumetric estimation and density characterization for operational MST systems.