Project: As-Built Modeling
This research thread seeks to build faithful models of physical
objects observed in the real world. By focusing on physical objects,
the research seeks to differentiate from standard
Structure-from-Motion or SLAM techniques through the exploiting the
compact nature of the object's support. Any additional geometry or
information that is available is also used.
Application of SfM/SLAM to civil engineering is motivated by the grand
challenges of engineering as reported by the National Academy of
Engineering (2008) in the document "Restoring and Improving Urban
Infrastructure." Modeling and annotating the as-built nature of
existing infrastructure is a time consuming task. While it may seem
beneficial to avoid this task, doing so leads to incomplete quality
assessment during construction and increases maintenance costs
downstream. Establishing an automated system for generating as-built
building information models (BIM) from visual measurements of as-built
infrastructure would eliminate the current bottleneck and enable more
complete and efficient operations.
Research in this area has included the investigation of SfM/SLAM
techniques, their accuracy, and their modification to suite civil
engineering needs and practice. Current research seeks to move from
point cloud data to BIM automatically via machine learning for
cost-effective generation of as-built models.
This research was supported in part by the National Science Foundation
Any opinions, findings, and conclusions or recommendations expressed in
this material are those of the author(s) and do not necessarily reflect
the views of the National Science Foundation.