• Citer

Automated Framework for Meshing of 3D Realistic Masonry Multi-Leaf Volumes from Wall Images

par Hermans, Yahroun Fei ;Ehab Moustafa Kamel, Karim ;Milani, Gabriele;Massart, Thierry Jacques
Référence World Congress on Computational Mechanics (WCCM)(17: 19/072026-24/7/2026: Munich), Proceedings of 17th World Congress on Computational Mechanics (WCCM)
Publication Publié, 2026-07-24

Abstract de conférence

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Accurate numerical modelling of unreinforced historical masonry structures requires geometric representations capable of capturing both global block arrangements and local stone irregularities. This contribution presents an automated framework for generating 3D synthetic masonry volumes from a single photograph of a wall surface, suitable for numerical simulations.A 2D segmentation of the input photograph is first performed to obtain labelled surface blocks for the visible outer leaf. Depending on the modelling assumptions, this surface layout can be (i) replicated to the opposite outer leaf (symmetric wall faces) or (ii) treated independently by generating a distinct surface layout for the second leaf, allowing nonmatching block patterns. These are consequently expanded using a watershed-based approach, lumping the joints, and then extruded to define the outer stone layer. The resulting volume is voxelized to establish a discretized inner domain, within which a constrained tessellation is generated from randomly sampled seeds with controlled inter-seed spacing.For each tessellation cell, a signed distance field is computed on the voxel grid and perturbed by a structured random noise field designed to reproduce cutting planes and natural surface roughness. Thresholding the combined field yields irregular, potentially non-convex synthetic stones while retaining global control over spacing through the initial constrained seeding. Surface and internal stones are assembled into a single container volume with interpenetration prevention. A global signed distance field is then processed using a weighted 3D watershed, mitigating staircase artefacts commonly associated with grid-based representations and enabling high-quality meshing. The final geometry is meshed using CGAL, and their use is illustrated through 3D limit analysis simulations.