Decay assessment in historic buildings with texture-based classification techniques: from digital survey to HBIM

This research work focuses on innovative classification strategies for the diagnosis of degradation in the field of Cultural Heritage. From the analysis of recent literature, the vital role that digital technologies, such as photogrammetry and laser scanning, play in the preservation process of historical buildings emerges. These methods enable the collection of accurate 2D and 3D data to assess buildings and their structural elements. Additionally, Virtual Reality offers cost-effective alternatives to traditional documentation methods. The research aims at leveraging photogrammetric techniques in the digital survey phase to acquire geometric and colorimetric data of heritage objects and panoramic views for virtual representation. Digital models derived from this data include spherical photos, point clouds, meshes and textures, which are then used to evaluate the degradation of the surrounding environment. The methodology is based on semi-automatic classification techniques to semantically segment and evaluate these degradation phenomena. The field of automatic classification of cultural heritage lacks specialized approaches, unlike other fields where it is highly developed. The research highlights the absence of standardized procedures for the acquisition, reconstruction and analysis of 2D/3D data in the evaluation of heritage artefacts, hence the need for streamlined, accessible and rapid methodologies to evaluate heritage conditions. To address these challenges, the research introduces preliminary results that aim to develop algorithms for monitoring the surfaces of historic buildings. This systematic workflow uses cost-effective, intuitive tools and machine learning algorithms to automatically recognize various types of decay in images, photos and textures resulting from monitoring and diagnostic activities. Finally, the paper emphasizes the potential of Historic Building Information Modeling to integrate and share different data sources, supporting restoration and analysis efforts while providing a valuable tool for project stakeholders and users. The proposed methodology is presented through an application in a case study.