Knowledge-Intensive 3D Reconstruction for Ancient Chinese Architecture

A groundbreaking approach to ancient architectural heritage preservation achieves unprecedented precision and semantic richness, bridging the gap between geometric accuracy and historical knowledge. The study introduces a dual-loop collaborative mechanism that synchronizes high-precision 3D reconstruction with semantic enrichment using LiDAR, UAV imagery, and historical maps, achieving 1.2 cm geometric accuracy and 97% semantic completeness across multiple sites. By leveraging CIDOC CRM-extended ontologies and JSON-LD standards, the framework encodes component hierarchies, decorative patterns, and color pedigrees, enabling traceable knowledge representation. A hybrid U-former and meta-learning model further enhances reconstruction by learning from cross-site morphological priors, while t-SNE clustering quantifies stylistic evolution from Song to Qing dynasties. Validated at Yingxian Wooden Pagoda, Foguang Temple, and Chengde Mountain Resort, this method demonstrates effective knowledge reuse across heterogeneous heritage sites.

Liu, S., & Jin, X. (2026). Knowledge-Intensive 3D Reconstruction for Ancient Chinese Architecture. International Journal of Knowledge Management, 22(1), 1–21. https://doi.org/10.4018/ijkm.416288