Automated 3D wireframe modeling of indoor structures from point clouds using constrained least-squares adjustment for as-built BIM

Jaehoon Jung, Sungchul Hong, Sanghyun Yoon, Jeonghyun Kim, Joon Heo

Research output: Contribution to journalArticle

24 Citations (Scopus)

Abstract

Automated three-dimensional (3D) modeling of building interiors for an as-built building information model (BIM) incurs critical difficulties because of the complex design of indoor structures and a variety of clutter from scanned point clouds. This paper proposes a scheme for automated 3D geometric modeling of indoor structures, including detailed components such as windows and open doors. Moreover, to produce a regularized model, we imposed constrained least-squares adjustment according to an assumption made with respect to typical indoor structures: walls, ceiling, floor, doors, and windows are composed of straight lines, either parallel or orthogonal. Modeling proceeds in two main phases: wall modeling and window modeling. In the wall modeling phase, the point-cloud acquisitions are projected onto a 2D binary image and the wall boundary is traced out. The boundary is regularized by means of the constrained least-squares method, and, thereby, a 2D floor boundary map can be obtained that later produces, by incorporation of a RANdom sample consensus-based height estimation, a 3D wall model. In the window modeling phase, the wall points are each projected onto a 2D inverse binary image to detect the hollow parts, after which image-based filtering removes incorrect detections. The lines, representing the four window edges, are extracted and adjusted again using the constrained least-squares method. Finally, the window models are projected back onto the original coordinates, and the complete 3D wireframe model is obtained. Two typical indoor environments, a seminar room and a corridor, both at Yonsei University, were chosen as study sites. The generated 3D wireframe models were evaluated by comparison with total station measurements. The 3D positional accuracies of the two results were 0.068 and 0.059 m, respectively, at the 90% confidence level, which indicates the effectiveness of the proposed approach. Finally, using the automatically produced wireframe model as a reference, as-built BIM creation was conducted in the BIM software.

Original languageEnglish
Article number04015074
JournalJournal of Computing in Civil Engineering
Volume30
Issue number4
DOIs
Publication statusPublished - 2016 Jul 1

Fingerprint

Binary images
Ceilings
Technical presentations

All Science Journal Classification (ASJC) codes

  • Civil and Structural Engineering
  • Computer Science Applications

Cite this

@article{6d1dc082a6aa4dc19241dad97a328496,
title = "Automated 3D wireframe modeling of indoor structures from point clouds using constrained least-squares adjustment for as-built BIM",
abstract = "Automated three-dimensional (3D) modeling of building interiors for an as-built building information model (BIM) incurs critical difficulties because of the complex design of indoor structures and a variety of clutter from scanned point clouds. This paper proposes a scheme for automated 3D geometric modeling of indoor structures, including detailed components such as windows and open doors. Moreover, to produce a regularized model, we imposed constrained least-squares adjustment according to an assumption made with respect to typical indoor structures: walls, ceiling, floor, doors, and windows are composed of straight lines, either parallel or orthogonal. Modeling proceeds in two main phases: wall modeling and window modeling. In the wall modeling phase, the point-cloud acquisitions are projected onto a 2D binary image and the wall boundary is traced out. The boundary is regularized by means of the constrained least-squares method, and, thereby, a 2D floor boundary map can be obtained that later produces, by incorporation of a RANdom sample consensus-based height estimation, a 3D wall model. In the window modeling phase, the wall points are each projected onto a 2D inverse binary image to detect the hollow parts, after which image-based filtering removes incorrect detections. The lines, representing the four window edges, are extracted and adjusted again using the constrained least-squares method. Finally, the window models are projected back onto the original coordinates, and the complete 3D wireframe model is obtained. Two typical indoor environments, a seminar room and a corridor, both at Yonsei University, were chosen as study sites. The generated 3D wireframe models were evaluated by comparison with total station measurements. The 3D positional accuracies of the two results were 0.068 and 0.059 m, respectively, at the 90{\%} confidence level, which indicates the effectiveness of the proposed approach. Finally, using the automatically produced wireframe model as a reference, as-built BIM creation was conducted in the BIM software.",
author = "Jaehoon Jung and Sungchul Hong and Sanghyun Yoon and Jeonghyun Kim and Joon Heo",
year = "2016",
month = "7",
day = "1",
doi = "10.1061/(ASCE)CP.1943-5487.0000556",
language = "English",
volume = "30",
journal = "Journal of Computing in Civil Engineering",
issn = "0887-3801",
publisher = "American Society of Civil Engineers (ASCE)",
number = "4",

}

Automated 3D wireframe modeling of indoor structures from point clouds using constrained least-squares adjustment for as-built BIM. / Jung, Jaehoon; Hong, Sungchul; Yoon, Sanghyun; Kim, Jeonghyun; Heo, Joon.

In: Journal of Computing in Civil Engineering, Vol. 30, No. 4, 04015074, 01.07.2016.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Automated 3D wireframe modeling of indoor structures from point clouds using constrained least-squares adjustment for as-built BIM

AU - Jung, Jaehoon

AU - Hong, Sungchul

AU - Yoon, Sanghyun

AU - Kim, Jeonghyun

AU - Heo, Joon

PY - 2016/7/1

Y1 - 2016/7/1

N2 - Automated three-dimensional (3D) modeling of building interiors for an as-built building information model (BIM) incurs critical difficulties because of the complex design of indoor structures and a variety of clutter from scanned point clouds. This paper proposes a scheme for automated 3D geometric modeling of indoor structures, including detailed components such as windows and open doors. Moreover, to produce a regularized model, we imposed constrained least-squares adjustment according to an assumption made with respect to typical indoor structures: walls, ceiling, floor, doors, and windows are composed of straight lines, either parallel or orthogonal. Modeling proceeds in two main phases: wall modeling and window modeling. In the wall modeling phase, the point-cloud acquisitions are projected onto a 2D binary image and the wall boundary is traced out. The boundary is regularized by means of the constrained least-squares method, and, thereby, a 2D floor boundary map can be obtained that later produces, by incorporation of a RANdom sample consensus-based height estimation, a 3D wall model. In the window modeling phase, the wall points are each projected onto a 2D inverse binary image to detect the hollow parts, after which image-based filtering removes incorrect detections. The lines, representing the four window edges, are extracted and adjusted again using the constrained least-squares method. Finally, the window models are projected back onto the original coordinates, and the complete 3D wireframe model is obtained. Two typical indoor environments, a seminar room and a corridor, both at Yonsei University, were chosen as study sites. The generated 3D wireframe models were evaluated by comparison with total station measurements. The 3D positional accuracies of the two results were 0.068 and 0.059 m, respectively, at the 90% confidence level, which indicates the effectiveness of the proposed approach. Finally, using the automatically produced wireframe model as a reference, as-built BIM creation was conducted in the BIM software.

AB - Automated three-dimensional (3D) modeling of building interiors for an as-built building information model (BIM) incurs critical difficulties because of the complex design of indoor structures and a variety of clutter from scanned point clouds. This paper proposes a scheme for automated 3D geometric modeling of indoor structures, including detailed components such as windows and open doors. Moreover, to produce a regularized model, we imposed constrained least-squares adjustment according to an assumption made with respect to typical indoor structures: walls, ceiling, floor, doors, and windows are composed of straight lines, either parallel or orthogonal. Modeling proceeds in two main phases: wall modeling and window modeling. In the wall modeling phase, the point-cloud acquisitions are projected onto a 2D binary image and the wall boundary is traced out. The boundary is regularized by means of the constrained least-squares method, and, thereby, a 2D floor boundary map can be obtained that later produces, by incorporation of a RANdom sample consensus-based height estimation, a 3D wall model. In the window modeling phase, the wall points are each projected onto a 2D inverse binary image to detect the hollow parts, after which image-based filtering removes incorrect detections. The lines, representing the four window edges, are extracted and adjusted again using the constrained least-squares method. Finally, the window models are projected back onto the original coordinates, and the complete 3D wireframe model is obtained. Two typical indoor environments, a seminar room and a corridor, both at Yonsei University, were chosen as study sites. The generated 3D wireframe models were evaluated by comparison with total station measurements. The 3D positional accuracies of the two results were 0.068 and 0.059 m, respectively, at the 90% confidence level, which indicates the effectiveness of the proposed approach. Finally, using the automatically produced wireframe model as a reference, as-built BIM creation was conducted in the BIM software.

UR - http://www.scopus.com/inward/record.url?scp=84975317723&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84975317723&partnerID=8YFLogxK

U2 - 10.1061/(ASCE)CP.1943-5487.0000556

DO - 10.1061/(ASCE)CP.1943-5487.0000556

M3 - Article

AN - SCOPUS:84975317723

VL - 30

JO - Journal of Computing in Civil Engineering

JF - Journal of Computing in Civil Engineering

SN - 0887-3801

IS - 4

M1 - 04015074

ER -