In the digital era, with the emergence of advanced reality-capture technologies, there is a growing recognition of the importance of documentation and monitoring of architectural heritage (AH) in conservation efforts. Techniques such as terrestrial laser scanning (TLS), mobile LiDAR, and photogrammetry generate high-resolution point clouds (PCs), providing detailed geometric and radiometric data for heritage structures. However, the complexity and unstructured nature of PCs can pose challenges related to data processing, segmentation, and storage.
One possible step in the right direction could be the identification of relevant geometric attributes, through a process of feature extraction, with the aim of optimising segmentation and classification processes. Segmentation, i.e. the subdivision of PCs into coherent subsets, could allow the efficient analysis, interpretation and visualisation of architectural elements. This process could in turn enhance the application of Heritage-Building Information Modeling (HBIM), degradation mapping and structural assessments, facilitating non-invasive monitoring and long-term conservation strategies.
Although the scientific community emphasises the use of artificial intelligence solutions and algorithms to automate processes, this contribution proposes instead the development of structured workflows for processing PCs, analysing the potential of geometric features in order to enrich 3D models with semantic information. The study is part of the Tech4You project within the National Recovery and Resilience Plan (PNRR).
