Application of Graph Neural Network in Matching Intangible Cultural Heritage Models with Virtual Reality Scenes

Abstract

This study aims to address the matching challenge between the computer-aided design (CAD) model (ICH) and the virtual reality (VR) scene, ultimately enhancing the digital preservation and presentation of ICH. We introduce an innovative matching approach leveraging the graph neural network (GNN) to accomplish this. Initially, by constructing a GNN model, we capture the spatial structure and relationships within the CAD model, extracting crucial features. Subsequently, these extracted features facilitate matching with corresponding elements in the VR environment. Our findings reveal that GNN exhibits significant advantages in aligning the CAD model of ICH with the VR scene, thereby effectively boosting the digital conservation and showcase of ICH. Experimental outcomes confirm the superiority of the GNN-based matching technique in this context. This study offers a novel technological approach for digital safeguarding and exhibition of ICH, which has substantial practical and application value.