Abstract

This thesis presents a comprehensive pipeline for three-dimensional (3D) reconstruction utilizing an innovative integration of augmented reality (AR) scanning, cloud-based processing, and multi-platform rendering, including virtual reality (VR) and traditional web-based environments. Aimed at improving the fidelity and usability of 3D reconstructions in industrial applications, the pipeline harnesses the synergy between a Unity-based AR scanning application and Gaussian Splatting algorithms executed in the cloud to generate detailed 3D models from two-dimensional (2D) images. The models can subsequently be rendered in VR and Unity for interactive user engagement. The AR scanning application guides users through the image capture process, ensuring optimal coverage and quality for reconstruction. The captured images are processed using Gaussian Splatting, a method chosen for its effectiveness in creating high-quality point clouds, which form the basis for the reconstructed models. The rendering aspect of the pipeline allows for flexible user interaction with the 3D models, both within an immersive VR environment and through standard web browsers, catering to various use cases and user preferences. The cloud pipeline ensures that the computationally intensive tasks are offloaded, facilitating scalability and access across diverse platforms. The work contributes to the field of 3D reconstruction by detailing the development process of each component, providing a thorough analysis of the system’s performance, and offering insights into user interactions with the technology across different rendering platforms. The findings indicate that the pipeline achieves high-quality 3D reconstructions with potential for application in a wide range of industrial contexts.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]