A novel methodology for the path alignment of visual SLAM in indoor construction inspection

Tao Lu*, Sonja Tervola, Xiaoshu Lü, Charles J. Kibert, Qunli Zhang, Tong Li, Zhitong Yao

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

2 Citations (Scopus)

Abstract

Path alignment is the process of mapping an indoor construction inspection path reconstructed by a visual SLAM system onto a 2D map with user interaction required to pinpoint at least two common tie points. In practice, more points are often needed due to path distortions and linear transformations, potentially resulting in reduced productivity. This paper proposes a methodology that combines two novel algorithms for the path alignment: (1) PCA_STAN_ALGO applies principal component analysis to remove path distortions caused by the xz plane of a camera coordinate system not being parallel to the floor plane; and (2) GRPX_TRANS utilizes a graphical user interface to facilitate the path alignment. The proposed methodology enables the users to utilize just two tie points for successful path alignment. An experimental study showed that applying both PCA_STAN_ALGO and GRPX_TRANS saved about 50% in time compared to using only GRPX_TRANS, a result of needing minimal moving points.

Original languageEnglish
Article number103723
Number of pages17
JournalAutomation in Construction
Volume127
DOIs
Publication statusPublished - Jul 2021
MoE publication typeA1 Journal article-refereed

Keywords

  • 2-Point Scheme
  • Affine transformation
  • Path alignment
  • Path distortion
  • Principal component analysis
  • Simultaneous localization and mapping

Fingerprint

Dive into the research topics of 'A novel methodology for the path alignment of visual SLAM in indoor construction inspection'. Together they form a unique fingerprint.

Cite this