Scholarlink

<p>Cylinders are prevalent structural units in pipelines, and their accurate and robust detection from 3D scanned point clouds is crucial for rapid reverse engineering. Existing methods often impose restrictions on cylinder parameters, limiting their applicability in complex pipeline scenarios. To address this, we propose a novel method for automatically detecting cylinders from unstructured point clouds. Our approach involves iterative clustering segmentation to reduce data complexity, reliable candidate cylinder estimation using three-point random sampling, high-precision cylinder fitting, and multi-filtering mechanisms to minimize false detections. Experimental results on both simulated and real-world data demonstrate that our method achieves precision, recall, and F1 scores of 0.8727, 0.8090, and 0.8397, respectively, outperforming existing methods. This work showcases the potential of our approach for automating the reverse engineering design of complex pipelines. Project Web: <a href="https://github.com/GCCao/Cylinders_detection_Cao_V2">https://github.com/GCCao/Cylinders_detection_Cao_V2</a>.</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

Automated detection of cylindrical structures in complex pipelines using iterative point cloud segmentation and high-precision fitting

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月27日

查看全文
  • Gengchen Cao

摘要

Cylinders are prevalent structural units in pipelines, and their accurate and robust detection from 3D scanned point clouds is crucial for rapid reverse engineering. Existing methods often impose restrictions on cylinder parameters, limiting their applicability in complex pipeline scenarios. To address this, we propose a novel method for automatically detecting cylinders from unstructured point clouds. Our approach involves iterative clustering segmentation to reduce data complexity, reliable candidate cylinder estimation using three-point random sampling, high-precision cylinder fitting, and multi-filtering mechanisms to minimize false detections. Experimental results on both simulated and real-world data demonstrate that our method achieves precision, recall, and F1 scores of 0.8727, 0.8090, and 0.8397, respectively, outperforming existing methods. This work showcases the potential of our approach for automating the reverse engineering design of complex pipelines. Project Web: https://github.com/GCCao/Cylinders_detection_Cao_V2.