Machine Learning Experiences for Information Enrichment of HBIM Models: the DECAI Project

In the management process of the architectural restoration site, we are witnessing the adoption of processes conducted largely in the traditional way increasingly alongside activities strongly influenced by new technologies. In particular, the new instruments for the metric survey return more and more information at very high resolution, progressively reducing the acquisition time. In recent years, In order to manage and structure these data, the disciplines of Representation and Geomatics have developed increasingly efficient restitution workflows by applying Building Information Modeling (BIM) processes to the built heritage (HBIM).  While the acquisition tools have evolved rapidly, the same cannot be said for the methodologies for managing the collected data and the tools for informed models. At the regard, there is a growing need to apply Machine Learning processes to improve data management and modelling in heritage restitution processes. 

The paper briefly presents the semi-automatic DECAI workflow for mapping architectural surfaces and their annotation in the HBIM environment. The case study is the elevation on Via Porta Praetoria, the facades of the hallway and the facades of the internal cavity of Palazzo Ansermin, in Aosta. The paper does not presume to replace the traditional survey procedures but rather to propose an integration and optimization, following a workflow that starts with a photogrammetric survey and ends with the informative enrichment of an HBIM model of the building.