Visualization of Cultural Heritage Models with Semantic Region-Level Narratives Using glTF Metadata Extensions

Abstract. This study presents a web-based workflow for linking predefined semantic regions of a 3D cultural heritage model with externally managed historical, material, architectural, and restoration-related records. Saint Michael Church in Akçaabat, Trabzon, was used as the test asset. Its photogrammetric model was converted to glTF and enriched with stable triangle identifiers through EXT_structural_metadata. The model was rendered in a Three.js-based viewer, while semantic and narrative records were maintained in MongoDB. When a user selected a surface, raycasting identified the intersected triangle and sent its identifier to the application programming interface. MongoDB then returned the semantic-region document whose children_ids array contained the selected triangle identifier. The retrieved information was displayed in the interface, and all triangles belonging to the matched region were highlighted on the model. Using one triangle identifier instead of three vertex identifiers reduced the identifier payload and required database references by approximately 66.7%. The prototype provided building-level access to architectural-plan and restoration-history records and region-level access to the large window, north entrance, apse, and nave dome. All tested components were successfully retrieved and highlighted, with displayed vertex counts ranging from 57,642 to 302,955. The results demonstrate the functional feasibility of connecting browser-based 3D interaction with separately managed and updateable cultural heritage narratives.