Low-cost and scalable framework for generating 3D digital cousins of irregular objects

Khan, Zeeshan, Pan, Yue, Wang, Dalei, Li, Haijiang ORCID: https://orcid.org/0000-0001-6326-8133 and Chen, Airong 2026. Low-cost and scalable framework for generating 3D digital cousins of irregular objects. Automation in Construction 190 , 107096. 10.1016/j.autcon.2026.107096

Full text not available from this repository.

Abstract

3D Modeling of irregular objects with common morphological traits and preserved specific variations is essential for visualization, simulation, and data-driven analysis in engineering. However, traditional methods such as X-ray computed tomography or multi-view-based photogrammetry are costly, labor-intensive, and difficult to scale. This paper proposes a low-cost and scalable framework, termed IODC-3D, to generate 3D digital cousins of irregular objects from limited 2D images, enabling the creation of ready-to-use 3D digital models that preserve morphology. The process combines StyleGAN3-based generative augmentation, segmentation with the Segment Anything Model(SAM), and transformer-based single-image reconstruction using TripoSR, resulting in watertight meshes for various irregular material types. Additionally, a quantitative index of cousinhood is introduced to evaluate the automation of generated particles. Large-scale experiments with up to 75,000 generated particles show consistent morphological accuracy (CFI = 0.64–0.69), reliable sample-level acceptance, and over 20 times faster reconstruction than traditional multi-view methods, making IODC-3D a viable solution for developing large-scale digital material libraries.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Schools > Engineering
Additional Information:	License information from Publisher: LICENSE 1: URL: http://creativecommons.org/licenses/by-nc-nd/4.0/, Start Date: 2028-06-20
Publisher:	Elsevier
ISSN:	0926-5805
Date of Acceptance:	5 June 2026
Last Modified:	01 Jul 2026 11:01
URI:	https://orca.cardiff.ac.uk/id/eprint/187843

Actions (repository staff only)

Edit Item