The automatic generation of semantically rich 3D indoor models from raw point cloud data is a topic of intensive research in recent years.
In architecture and engineering, as 3D designs are frequently not available, or significantly different from the as-is condition of a given building, technology for generating Building Information Models (BIMs) from observations is indeed essential.
At the same time, though, the reconstruction of 3D indoor models is yet a challenging task, where an accurate representation of the building geometry is at the basis of a correct generation of BIMs. Hence, cheap, intuitive and effective tools for automatically collecting the 3D geometry of indoor environments are increasingly necessary.
In this context, the use of range camera technology can offer significant advantages over traditional methods, thanks to their easiness of use, especially when not qualified users are involved or when a rapid intervention is required [1].
It is thus essential to find new methods, able to extract the information contained in the raw 3D models so easily collected by the range cameras. The aim of this work is precisely to extend a previously developed algorithm to automatically retrieve the complete geometric information of environments starting from the raw 3D models acquired by range cameras and the fusion of scans of different rooms into an overall BIM. Indeed, currently the scan to BIM process, the process of converting 3D models into a BIM, is primarily a manual operation, at least in some of its parts, labor-intensive and error-prone [2].
[1] Capocchiano, F., Ravanelli, R., and Crespi, M. (2017). A Tool for Crowdsourced Building Information Modeling Through Low-Cost Range Camera: Preliminary Demonstration and Potential. International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences
[2] Xiong et al. (2013). Automatic creation of semantically rich 3D building models from laser scanner data. Automation in Construction
Despite several methods have been developed (Tang et al., 2010) for automatizing the scan-to-BIM process, nowadays none seems specifically designed to leverage the peculiarities of the range cameras. As explained above, range cameras are cheap, intuitive and effective tools for automatically collecting the 3D geometry of indoor environments, enabling their use also to not-expert operators in the context of crowdsourced VGI initiatives.
Thus, the present research will study in detail the application of range cameras to indoor modelling, paying specific attention to the BIM generation from their point clouds. A new version of a scan-to-BIM algorithm will be implemented, allowing the retrieval of the complete geometric information (e.g., elements such as the height of the walls, floors and ceilings) from point cloud data and the fusion of different rooms into in an overall BIM.