Abstract
Secondary building assets management requires a large amount of information related to them. Nevertheless, building assets surveys
are cost and time demanding, especially because they need long post-processing efforts in order to systematize collected data.
Furthermore, with the recent transition towards the BIM methodology for building management also modeling building objects
both in their geometric features and in their related information is a long process and error-prone task. Under these circumstances
the possibility of performing the majority of operation on-site would definitely make the process more efficient and it would reduce
errors. Augmented Reality (AR) with its capability of overlapping digital data to the real scene is the right tool to support operators
on-site.
The proposed system has the aim of reducing the time of secondary building assets survey and provide a tool for the automatic
enrichment of BIM models. An AR device (Hololens) with an embedded computer and a neural compute stick constitute the portable
on-site system for the automatic recognition of assets objects, removing the necessity of reworking data off site. A trained Deep
Learning Neural Network inside the neural compute stick performs the recognition providing the operator with objects features and
position. The AR application inside the Hololens operates as an interface between the user and the digital information just created.
Finally, data is stored in a NoSQL database linked to the BIM model so as to be available for further operations. The visually
supporting information provided by the AR tool, the possibility of working on data directly on site and the portability of the system
represent means for increasing efficiency in survey operations. First tests have been conducted so as to prove the feasibility of the
system and its use on site without further equipment.
are cost and time demanding, especially because they need long post-processing efforts in order to systematize collected data.
Furthermore, with the recent transition towards the BIM methodology for building management also modeling building objects
both in their geometric features and in their related information is a long process and error-prone task. Under these circumstances
the possibility of performing the majority of operation on-site would definitely make the process more efficient and it would reduce
errors. Augmented Reality (AR) with its capability of overlapping digital data to the real scene is the right tool to support operators
on-site.
The proposed system has the aim of reducing the time of secondary building assets survey and provide a tool for the automatic
enrichment of BIM models. An AR device (Hololens) with an embedded computer and a neural compute stick constitute the portable
on-site system for the automatic recognition of assets objects, removing the necessity of reworking data off site. A trained Deep
Learning Neural Network inside the neural compute stick performs the recognition providing the operator with objects features and
position. The AR application inside the Hololens operates as an interface between the user and the digital information just created.
Finally, data is stored in a NoSQL database linked to the BIM model so as to be available for further operations. The visually
supporting information provided by the AR tool, the possibility of working on data directly on site and the portability of the system
represent means for increasing efficiency in survey operations. First tests have been conducted so as to prove the feasibility of the
system and its use on site without further equipment.
Original language | English |
---|---|
Title of host publication | Proceedings of the Creative Construction Conference (CCC) |
Subtitle of host publication | CCC2019 |
Place of Publication | Budapest |
Pages | 806-811 |
Number of pages | 6 |
DOIs | |
Publication status | Published - 2019 |