Cloud-based Fusion Rendering for AR Applications

Jing Tang,Yuxi Zhou,Zhenyu Chen,Cheng Chen,Gang Xiong,Hongxin Zhang

DOI: https://doi.org/10.1109/dtpi59677.2023.10365411

2023-01-01

Abstract:To create digital twins, it is necessary to represent real objects or scenes in a virtual form. With the rapid development of Augmented Reality (AR) technology, its applications in various fields are becoming increasingly extensive. However, in traditional AR applications, the client has high performance requirements for device compute-intensive tasks, which cannot be met on low-performance devices. To solve this issue, in this paper, we propose and implement a novel digital twin method for AR technology, which is based on a cloud fusion platform. The approach includes deploying the Neural Radiance Field (NeRF) on a cloud server for high-quality model rendering and integrating simultaneous localization and mapping (SLAM) with AR visualization on the client side. This design effectively separates compute-intensive tasks from client-side rendering tasks, reducing the performance requirements of client devices and enhancing overall scalability and availability of the system. Practical application and tests show that the proposed AR system based on the cloud fusion platform performs well in terms of performance, stability, and scalability. This paper provides detailed discussions on the relevant technical methods and approaches, and showcases examples that serve as a foundation for the implementation of digital twins across diverse fields.