A comparative evaluation of 3D Gaussian splatting and LiDAR for digitizing handheld objects in VR

Abstract

This paper compares the effectiveness of mobile device-based LiDAR and 3D Gaussian Splatting for digitizing small objects for interactable use in immersive VR applications. The study aims to determine the most accurate and efficient consumer-accessible method for translating handheld real-world objects into immersive virtual environments using mobile phones with LiDAR sensors. Using a Meta Quest 2 controller as the test object-chosen for its complex geometry with a torus-like ring and a genus of 1, a combination of curved and flat surfaces, and interactable buttons-we conducted multiple mesh and 3D Gaussian Splatting (3D-GS) scans using two mobile apps: SiteScape for LiDAR-based point clouds, and Scaniverse for both mesh and Splat-based scans. The dimensional accuracy of each scanning technique was evaluated by comparing cloud-to-mesh distance of the generated models in the CloudCompare software against a highly accurate reference controller model from Meta, averaging the results of three models for each method to account for environmental variability. The models are then imported into Unreal Engine for a VR application to assess the controller's visual quality and interaction fidelity. The findings of this study provide a comparative analysis of the strengths and weaknesses of each method, offering insights into the most optimal mobile 3D scanning techniques for users to digitize small objects for interactive and immersive VR applications.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata

Altmetrics