Scholarlink

<p>Accurate rock fragmentation analysis is paramount for optimizing blasting efficiency and safety in mining operations. Common RGB-based methods for fragmentation analysis, including 2D images, orthophotos, and photogrammetry, are severely hampered by inconsistent lighting and shadowing, a common challenge in underground mines and night-shift operations. Such dependency compromises the reliability of particle size distribution, leading to suboptimal operational decisions. This study presents a comparative analysis of rock fragmentation measurement performance between 2D photogrammetry and LiDAR-based 3D point cloud approaches under low-light conditions. We applied our previously published LiDAR-based point cloud segmentation framework and compared it against a 2D photogrammetry-based deep learning approach on a blasted limestone rock pile (particle sizes 15–35 cm) across seven distinct scenarios: one optimal lighting condition (~ 250 lx) and six challenging scenarios replicating underground mine and night-shift conditions, including extreme darkness to dim lighting (0–15 lx), natural and artificial illumination, and varying shadow intensities and directions. Rock size distributions were analyzed using the 3D LiDAR method, which utilizes geometric coordinate data, and the 2D photogrammetry method processed through Agisoft Metashape. Performance was evaluated through characteristic size error at 20%, 50%, 80% and maximum percent passing (i.e. x20, x50, x80, xmax) equivalent diameter measurements. The 3D LiDAR method maintained a low error of 7–10% across all lighting conditions. In contrast, the 2D photogrammetry method’s accuracy proved highly dependent on lighting and shadows, achieving error of 2–17% under optimal lighting, degrading to 7–34% error under artificial lighting with shadows, and completely failing in darkness or partial illumination (0–10 lx). This underscores LiDAR’s critical advantage in capturing 3D rock geometries based on geometric features rather than RGB data, supporting consistent and reliable autonomous mining operations independent of lighting conditions.</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

Rock Fragmentation and Particle Size Distribution Analysis: A Comparative Study of Image-Based Photogrammetry and LiDAR Point Cloud Analysis

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • Mojgan Faramarzi H.,
  • Kamran Esmaeili,
  • Thomas Bamford

摘要

Accurate rock fragmentation analysis is paramount for optimizing blasting efficiency and safety in mining operations. Common RGB-based methods for fragmentation analysis, including 2D images, orthophotos, and photogrammetry, are severely hampered by inconsistent lighting and shadowing, a common challenge in underground mines and night-shift operations. Such dependency compromises the reliability of particle size distribution, leading to suboptimal operational decisions. This study presents a comparative analysis of rock fragmentation measurement performance between 2D photogrammetry and LiDAR-based 3D point cloud approaches under low-light conditions. We applied our previously published LiDAR-based point cloud segmentation framework and compared it against a 2D photogrammetry-based deep learning approach on a blasted limestone rock pile (particle sizes 15–35 cm) across seven distinct scenarios: one optimal lighting condition (~ 250 lx) and six challenging scenarios replicating underground mine and night-shift conditions, including extreme darkness to dim lighting (0–15 lx), natural and artificial illumination, and varying shadow intensities and directions. Rock size distributions were analyzed using the 3D LiDAR method, which utilizes geometric coordinate data, and the 2D photogrammetry method processed through Agisoft Metashape. Performance was evaluated through characteristic size error at 20%, 50%, 80% and maximum percent passing (i.e. x20, x50, x80, xmax) equivalent diameter measurements. The 3D LiDAR method maintained a low error of 7–10% across all lighting conditions. In contrast, the 2D photogrammetry method’s accuracy proved highly dependent on lighting and shadows, achieving error of 2–17% under optimal lighting, degrading to 7–34% error under artificial lighting with shadows, and completely failing in darkness or partial illumination (0–10 lx). This underscores LiDAR’s critical advantage in capturing 3D rock geometries based on geometric features rather than RGB data, supporting consistent and reliable autonomous mining operations independent of lighting conditions.