TY - GEN
T1 - Decentralized visual 3d mapping of scattered work locations for high-frequency tracking of indoor construction activities
AU - Amer, Fouad
AU - Golparvar-Fard, Mani
N1 - Publisher Copyright:
© 2018 American Society of Civil Engineers (ASCE). All rights reserved.
PY - 2018
Y1 - 2018
N2 - Complete 3D mapping of indoor construction scenes using images and videos is a challenging task. It requires all workspaces to be continuously captured to enable complete modeling of the rapidly changing environments. Nonetheless, manually mapping an entire workspace on a daily basis is time-consuming. Also, at building sites, guaranteeing complete 3D mapping require carefully accounting for transitions in hallways for moving the camera from one workspace to another. To address these challenges, this paper presents a new method together with a workflow that optimizes data capture and enables complete 3D modeling. Using daily images taken, a pipeline of structure from motion coupled with multi-view dense reconstruction algorithms enables complete 3D modeling in each workspace. Walkthrough videos are then recorded weekly and are provided as input to a variant of a SLAM algorithm to tie in locally reconstructed workspaces in 3D. This enables project teams to capture images of each workspace-where work is continuously changing-at a high frequency (e.g., daily) and continuously produce dense 3D models. It also leverages a video capture of the entire workspace on a low frequency (e.g., weekly) to tie in locally produced dense 3D models. Preliminary results from a project site show that by decentralizing the image capture of each workspace to the trades as part of their required daily field reporting, and rapidly videotaping all work locations by a site engineer as part of a weekly walkthrough, the entirety of an indoor phase of construction project can be mapped in 3D. The specifics of the method and the benefits and challenges of the implementation workflow are discussed in detail.
AB - Complete 3D mapping of indoor construction scenes using images and videos is a challenging task. It requires all workspaces to be continuously captured to enable complete modeling of the rapidly changing environments. Nonetheless, manually mapping an entire workspace on a daily basis is time-consuming. Also, at building sites, guaranteeing complete 3D mapping require carefully accounting for transitions in hallways for moving the camera from one workspace to another. To address these challenges, this paper presents a new method together with a workflow that optimizes data capture and enables complete 3D modeling. Using daily images taken, a pipeline of structure from motion coupled with multi-view dense reconstruction algorithms enables complete 3D modeling in each workspace. Walkthrough videos are then recorded weekly and are provided as input to a variant of a SLAM algorithm to tie in locally reconstructed workspaces in 3D. This enables project teams to capture images of each workspace-where work is continuously changing-at a high frequency (e.g., daily) and continuously produce dense 3D models. It also leverages a video capture of the entire workspace on a low frequency (e.g., weekly) to tie in locally produced dense 3D models. Preliminary results from a project site show that by decentralizing the image capture of each workspace to the trades as part of their required daily field reporting, and rapidly videotaping all work locations by a site engineer as part of a weekly walkthrough, the entirety of an indoor phase of construction project can be mapped in 3D. The specifics of the method and the benefits and challenges of the implementation workflow are discussed in detail.
UR - http://www.scopus.com/inward/record.url?scp=85048622962&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85048622962&partnerID=8YFLogxK
U2 - 10.1061/9780784481264.048
DO - 10.1061/9780784481264.048
M3 - Conference contribution
AN - SCOPUS:85048622962
T3 - Construction Research Congress 2018: Construction Information Technology - Selected Papers from the Construction Research Congress 2018
SP - 491
EP - 500
BT - Construction Research Congress 2018
A2 - Wang, Chao
A2 - Berryman, Charles
A2 - Harris, Rebecca
A2 - Harper, Christofer
A2 - Lee, Yongcheol
PB - American Society of Civil Engineers
T2 - Construction Research Congress 2018: Construction Information Technology, CRC 2018
Y2 - 2 April 2018 through 4 April 2018
ER -