This study aims to design a "Korean-specific Ontology" based on CRMdig for the provenance (lifecycle) of Korean cultural heritage 3D data, and to construct semantic data through this framework. The goal is to integratedly manage 3D data, which has been underutilized by domestic institutions, mitigate information distortion (hallucination) phenomena occurring in the AI era, and further establish a foundation for international data linkage and trust-based services.

Entering the era of Artificial Intelligence (AI), the value of digital data is rising to unprecedented levels. In particular, the Korean cultural heritage sector has established high-quality 3D original record data generated through 3D scanning and photogrammetry technologies as core assets, following the "Digital Heritage Transformation 2030" plan announced in 2021(Korea Heritage Service 2021), and this effort is currently actively ongoing.

However, cultural heritage 3D data construction projects, conducted with massive resource investment, have focused on visual representation—specifically the "Digital Twin"—while neglecting the management of "Digital Provenance," which tracks the context of data generation and technical history. This lack of transparency is a major cause of degrading the academic reliability of data and hindering its secondary use (reuse). Accordingly, it is an urgent task not only to provide high-quality 3D data or unprocessed raw data but also to design a standardized 3D data ontology that can effectively integrate data, transparently manage its source and history (provenance), and guarantee international interoperability in preparation for all future stages of data disclosure and utilization.

Therefore, this study intends to design a "Korean Cultural Heritage 3D Data Lifecycle Ontology" by adopting CIDOC-CRM, the ICOM cultural heritage standard ontology, and its digital extension model, CRMdig, as core methodologies (International Committee for Documentation n.d.; CIDOC CRM Special Interest Group n.d.).

This research proceeds in a total of five stages.

First, we analyze the overall lifecycle of domestic cultural heritage 3D data. We closely investigate the status of raw data acquisition, data post-processing, final deliverables, and data archiving, centering on the Korea Heritage Service (KHS) and the National Museum of Korea (NMK), which represent national heritage(Korea Heritage Service n.d.; National Museum of Korea n.d.).  Although both institutions record precise metadata based on detailed data construction guidelines, they demonstrate the fundamental limitations of archiving systems based on traditional database structures where users search for information via keywords.

Second, we explore standard models through the analysis of overseas 3D archive cases. We investigate the level of data provenance management in major overseas 3D archives, centering on four leading institutions: Europeana, the Smithsonian, the British Museum, and CyArk, and perform a comparative analysis with the domestic situation(Europeana n.d.; Smithsonian Institution n.d.; The British Museum n.d.; OpenHeritage3D n.d.) . Through these cases, we identify how data schemas are designed differently according to the core objectives of each archive. Synthesizing these domestic and international conditions, we determine the direction for this study.

Third, based on the previous analysis results, we design a "Korean Cultural Heritage 3D Data Ontology" based on the international standard semantic model, CRMdig. In this stage, we focus particularly on the expression of provenance (data history) that can track the entire process from generation to processing and output of 3D data. While CRMdig is an extension model specialized to precisely capture the dynamic lifecycle of digital objects, it does not pre-define classes that encompass all detailed work steps of a specific domain. For example, the D2_Digitization_Process class of CRMdig expresses all processes of digitizing a target but has limitations in distinguishing specific work types such as "Terrestrial LiDAR Scan," "Structured Light Scan," and "Aerial Photogrammetry." To clearly express such technical details, we expand the ontology by proposing a user-defined classification system composed of sub-classes of D2_Digitization_Process or properties modifying it. To verify the effectiveness of the designed ontology, we select a specific cultural heritage asset as a case study and build a pilot dataset. In this process, we assign Unique Resource Identifiers (URIs) to all digital objects and transformation events, ranging from the acquired raw dataset to data processing techniques and final results, thereby clearly defining the connection relationships of each stage(Figure 1).

Figure 1. Conceptual Framework of the Proposed CRMdig-based Provenance Ontology

Fourth, based on the designed ontology schema and the pilot dataset, we generate RDF (Resource Description Framework) triples, which are sets of interconnected data. These generated RDF triples are finally completed as a Knowledge Graph containing detailed provenance information of the corresponding cultural heritage.

Fifth, the utility of the constructed Knowledge Graph is finally verified through SPARQL query testing. This process evaluates whether the designed ontology has secured practical data traceability and interoperability by performing complex semantic queries that were impossible with conventional simple metadata searches. For instance, by successfully executing a query such as "Retrieve the lineage of all 3D data generated at Temple Site A during a specific period (2000–2025)," this study demonstrates the connective relationships between the final outcome (Final 3D Data) and the source data (Raw Data), centering on the history of the 3D data. Through this, we demonstrate that the proposed ontology can retrace the entire data lifecycle and is universally applicable to cultural heritage data of various scales.

Ultimately, this study will establish a long-term preservable 3D data system based on FAIR principles, thereby laying the foundation for international data linkage of Korean cultural heritage. In the future, this can lead to multifaceted research expansion, such as automating the structuring process of messy real-world data using AI and developing systems that link semantic data with LLM models.