A Semi-Automated Pipeline for Extracting Architectural Plans from 3D LiDAR Data of Ancient Heritage Sites

Marianna Bartrick-Krana, Roberto de Lima, Aziliz Vandesande, Maarten Bassier

Abstract. This paper presents a semi-automated pipeline for extracting architectural plans from terrestrial LiDAR point clouds of archaeological sites characterized by irregular geometries and significant surface degradation. The proposed workflow converts dense three-dimensional data into simplified two-dimensional representations by combining geometric alignment, expert-guided cross-section selection, feature detection using Intrinsic Shape Signatures (ISS), region-based segmentation, and polyline simplification. The method is specifically designed to handle rock-cut tomb environments, where erosion, noise, and non-planar surfaces complicate conventional Scan-to-Plan approaches. The pipeline is evaluated on six tombs from the Sheikh Said necropolis in Middle Egypt, covering a range of architectural configurations and preservation states. Results demonstrate that the method efficiently produces CAD-compatible outlines that capture the dominant structural geometry while significantly reducing manual drafting time. However, the accuracy of the generated plans depends on surface conditions, with degraded or ornamented areas introducing geometric artifacts that require expert refinement. The proposed approach provides a robust geometric baseline for archaeological documentation and highlights the potential of human-in-the-loop workflows for complex heritage environments.