GAUSSIAN SPLATTING FOR CULTURAL HERITAGE DIGITIZATION: TECHNICAL POSITIONING, APPLICATION SCENARIOS, AND FUTURE PROSPECTS

Abstract

Cultural heritage digitization is increasingly shifting from static geometric recording toward immersive, interactive, and scene-based representation. Within this technological transformation, Gaussian Splatting has emerged as a promising approach for image-driven three-dimensional visualization, offering photorealistic rendering, spatial continuity, and real-time free-viewpoint navigation. This article examines the role of Gaussian Splatting in cultural heritage digitization by situating it within the broader evolution from manual modeling, laser scanning, photogrammetry, and SfM-MVS reconstruction to neural rendering. It argues that Gaussian Splatting should be understood primarily as a rendering-oriented scene representation rather than as a substitute for measurement-oriented point clouds, mesh models, BIM/HBIM, or GIS-based heritage information systems. On this basis, the article discusses its potential applications in architectural heritage, archaeological sites, caves and grottoes, museum collections, virtual museums, and digital cultural tourism. It further analyzes key methodological stages, including data acquisition, pose estimation, Gaussian representation and optimization, quality evaluation, semantic segmentation, multi-source fusion, and platform deployment. While Gaussian Splatting provides strong advantages in visual realism and interactive access, its limitations in geometric measurability, semantic organization, long-term interoperability, and conservation-grade documentation must be carefully addressed. The article concludes that the future value of Gaussian Splatting lies not in replacing established digitization methods, but in connecting visual scenes with measurement data, semantic annotation, archival systems, and public-facing platforms, thereby contributing to more integrated, accessible, and sustainable cultural heritage digital infrastructures.