An integrated approach for modelling and global registration of point clouds

Tahir Rabbani*, Sander Dijkman, Frank van den Heuvel, George Vosselman

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

199 Citations (Scopus)

Abstract

Point cloud acquisition by using laser scanners provides an efficient way for 3D as-built modelling of industrial installations. Covering such an installation with point cloud data often requires data acquisition from multiple standpoints. Before the actual modelling can start the transformation parameters of all scans need to be determined. Two methods to register point clouds of industrial scenes with different coordinate definitions are presented. Corresponding object models in different scans are used to determine the translation and rotation parameters of the scans. The first method, called Indirect method, is a two-step approach as object fitting and registration of the scenes is done separately. The second method, called Direct method simultaneously determines the shape and pose parameters of the objects as well as the registration parameters. Both methods are designed such that optimal use can be made of the knowledge of shapes present in industrial environments. Compared to ICP the presented approach combines registration and modelling and thus avoids the accumulation of errors. Furthermore, the simultaneous registration of multiple scans is possible. The presented approaches are based on non-linear least squares and provide quality measures in the form of covariance matrix of the estimated parameters, which can be used to decide if more scans are needed, and how and where they should be captured. Results are presented on some point cloud data-sets from actual industrial sites, where registration was done without using any artificial targets.
Original languageEnglish
Pages (from-to)355-370
Number of pages16
JournalISPRS journal of photogrammetry and remote sensing
Volume61
Issue number6
DOIs
Publication statusPublished - 2007

Keywords

  • ADLIB-ART-2547
  • EOS
  • ITC-ISI-JOURNAL-ARTICLE

Fingerprint

Dive into the research topics of 'An integrated approach for modelling and global registration of point clouds'. Together they form a unique fingerprint.

Cite this