Concepts for Off-Site Anchoring of Digital Augmentations with On-Site Captured 3D Images
Year: 2025
Type: dissertation
Abstract: This is the main repository for the Master Thesis:Concepts for Off-Site Anchoring of Digital Augmentations with On-Site Captured 3D Images This repository contains the project with the code, a folder with the used 3D images and models, a video showcasing a demonstration, and the thesis PDF. The instructions on how to run the project are in the README.md of the project file. Abstract This work explores the integration of Extended Reality technologies into the field of modeling, focusing on the placement of digital models created off-site in their respective real-world environments on-site. While traditional modeling platforms like ADOxx operate in 2D environments, the M2AR modeling platform offers 3D modeling capabilities using WebXR and Three.js, enabling cross-device compatibility through web-based access. A key challenge addressed in this work is anchoring digital models to their corresponding real-world locations. To this end, three conceptual solutions were proposed. While Conceptual Solution 1 relies primarily on manual user input, Conceptual Solution 2 enhances this approach by integrating GNSS location data for improved guidance. Conceptual Solution 3 employs mostly automated approaches through the use of computer vision algorithms for location recognition and model placement. However, due to technical constraints, only Conceptual Solution 1 was implemented in this work. The implementation was tested across six diverse real-world locations and demonstrated functional usability for AR model placement using 3D images. While the approach proved effective in real-world environments with clear depth cues, it exhibited limitations in large open spaces or when dealing with flat or distant structures. 80% of the defined requirements were fully met. The solution’s capability to address the problem was successfully validated within the M2AR modeling platform.
Author: Odermatt, Lukas
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6893361402
Cites:
Cited by:
Related to: