Pipe spool recognition in cluttered point clouds using a curvature-based shape descriptor

Thomas Czerniawski, Mohammad Nahangi, Carl Haas, Scott Walbridge

Research output: Contribution to journalArticlepeer-review

58 Scopus citations


Automating dimensional compliance control and progress tracking using computer vision has been identified as a central opportunity for improvement in the construction industry. Current 3D imaging sensors provide massive amounts of spatial data that remain underutilized due to the prohibitively time-consuming manual process of extracting usable information. Desired information is typically centered on a specific object of interest within 3D images, so there is a need for construction specific object recognition processes. In this paper, we present an automated method for locating and extracting pipe spools in cluttered point cloud scans. The method is based on local data level curvature estimation, clustering, and bag-of-features matching. Experimental results from two point clouds containing pipe spool objects demonstrate the method's ability to successfully extract spools from cluttered scenes as well as differentiate between similar spools in a single scene.

Original languageEnglish (US)
Pages (from-to)346-358
Number of pages13
JournalAutomation in construction
Issue numberPart 2
StatePublished - Nov 1 2016
Externally publishedYes


  • 3D reconstruction
  • Bag-of-features
  • Building information models
  • Curvature
  • Object recognition
  • Pipe spool
  • Point clouds

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Civil and Structural Engineering
  • Building and Construction


Dive into the research topics of 'Pipe spool recognition in cluttered point clouds using a curvature-based shape descriptor'. Together they form a unique fingerprint.

Cite this