ENHANCING 3D MODELING EFFICIENCY VIA SEMI-AUTOMATIC POINT CLOUD SEGMENTATION AND MULTI-LOD MESH RECONSTRUCTION

The accuracy of 3D models in historical buildings is a debated topic due to the increasing demand for digital documentation and the need for automated post-processing methods to reduce manual labor and improve data analysis. The proposed method aims to improve 3D reconstruction efficiency for complex geometries exemplified by the Ognissanti Church of Trani (XII century), Italy. The methodology includes point cloud segmentation and classification algorithms, (RANSAC and Random Forest) to isolate architectural elements. The segmented portions undergo processing utilizing three 3D reconstruction algorithms: Alpha-Shape, Ball-Pivoting, and Poisson Surface Reconstruction. Customized settings enable polygonal meshes with varying levels of detail. Visual Programming Language operations refine the resulting meshes in terms of triangulation and computational efficiency, ensuring a high level of fidelity and applications in HBIM framework.