Automated Geometric Reconstruction of Degraded Masonry Structures from Point Cloud Data: An Uncertainty Quantification Framework for Heritage Building Assessment

The geometric characterization of masonry structures with incomplete or degraded surfaces remains a critical challenge in building rehabilitation, structural assessment, and BIM-based documentation. This study presents an automated computational framework for reconstructing the original geometry of partially collapsed masonry walls from Terrestrial Laser Scanning (TLS) data, providing quantitatively defined uncertainty bounds to support informed decision-making in heritage conservation and the built environment. The proposed methodology integrates four complementary analytical components: (i) automated wall thickness extraction based on PCA-driven point cloud segmentation combined with robust percentile statistics; (ii) wall trajectory reconstruction through the identification of polygonal configurations and structural discontinuities between segments; (iii) slope profile detection using local gradient analysis to characterize wall batter angles. (iv) maximum wall-height estimation through a limit-equilibrium model. The entire workflow is implemented in Python within Jupyter Notebooks, ensuring transparency, reproducibility, and compliance with FAIR principles. Material characterization, informed by archaeological evidence and geotechnical parameters, enables the estimation of structural height through a limit-equilibrium approach. A key contribution of this research is the Best-Fit Model Profile (BFMP) methodology for uncertainty quantification, which derives internal geometric references directly from the point cloud data. This approach allows for a systematic evaluation of preservation states and construction variability without reliance on external ground-truth measurements. Structural height estimation incorporates geotechnical and material parameters, whose influence is assessed through a sensitivity analysis, while the reported height interval propagates the thickness variability. Validation on a prehistoric masonry structure demonstrates the framework’s ability to transform ambiguous survey data into parametric, BIM-ready outputs with explicit uncertainty limits. Compared with traditional CAD-based reconstruction workflows, the proposed approach significantly reduces subjective interpretation while enhancing geometric reliability and analytical transparency.