Evaluation of Different Features for Matching Point Clouds to Building Information Models

Te Gao, Semiha Ergan, Burcu Akinci, James Garrett

Research output: Contribution to journalArticlepeer-review


With the increased usage of building information models (BIMs) during construction, has BIM become a medium for delivering as-built building information. It is important to maintain accurate and up-to-date information stored in a BIM so that it can become a reliable data source throughout the service life of a facility. Laser scanning technology is able to capture accurate geometric data in the form of a point cloud and to depict the existing condition of a building. Hence, point cloud data captured by laser scans can be used as references to update a given BIM. An important step during the update process is to match segments of elements captured by a point cloud to building components modeled in a BIM, so that the discrepancies between the two data sets can be identified. Typically, features depicted within point cloud segments and BIM components are used in the matching process. However, understanding is limited regarding which features enable the matching process and how these features perform. This paper describes six feature-based matching approaches that match segments of a point cloud to components modeled in a BIM. Next, it discusses the results of an experimental analysis conducted to evaluate the performance of different features used to match mechanical equipment and ductwork captured by point clouds to the corresponding objects modeled in an as-designed BIM.

Original languageEnglish (US)
Article number04014107
JournalJournal of Computing in Civil Engineering
Issue number1
StatePublished - Jan 1 2016


  • Building information model
  • Discrepancies
  • Matching
  • Point clouds
  • Precision
  • Recall

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Computer Science Applications


Dive into the research topics of 'Evaluation of Different Features for Matching Point Clouds to Building Information Models'. Together they form a unique fingerprint.

Cite this