A Joint Sensing Framework for Real-time Mobile Camera Localization
(2025) MAMM15 20251Department of Design Sciences
- Abstract
- In this paper, a joint perception framework combining fixed and mobile cameras is proposed, which can achieve accurate real-time positioning of the mobile camera in a unified coordinate system. The system integrates VGGT for high-precision pose estimation and combines it with MASt3r-SLAM for robust real-time tracking. Alignment between different coordinate systems is achieved through rotation matrix transformation and Procrustes analysis. User evaluation confirms the effectiveness of the system in indoor navigation scenarios and is expected to help visually impaired users. Although the framework is sensitive to motion blur, it shows strong flexibility, high accuracy, and effective use of multi-source image data. Future work will focus on... (More)
- In this paper, a joint perception framework combining fixed and mobile cameras is proposed, which can achieve accurate real-time positioning of the mobile camera in a unified coordinate system. The system integrates VGGT for high-precision pose estimation and combines it with MASt3r-SLAM for robust real-time tracking. Alignment between different coordinate systems is achieved through rotation matrix transformation and Procrustes analysis. User evaluation confirms the effectiveness of the system in indoor navigation scenarios and is expected to help visually impaired users. Although the framework is sensitive to motion blur, it shows strong flexibility, high accuracy, and effective use of multi-source image data. Future work will focus on enhancing robustness to blurred inputs, extending the framework to support 4D scene reconstruction, and developing advanced auxiliary navigation functions to pave the way for practical applications. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9193418
- author
- Zeng, Jiuming LU and Hua, Yanling LU
- supervisor
-
- Günter Alce LU
- organization
- course
- MAMM15 20251
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Real-time Camera Localization, Camera Pose Estimation, Joint Sensing, SLAM, 3D Foundation Models
- language
- English
- id
- 9193418
- date added to LUP
- 2025-06-05 12:08:47
- date last changed
- 2025-06-05 12:08:47
@misc{9193418, abstract = {{In this paper, a joint perception framework combining fixed and mobile cameras is proposed, which can achieve accurate real-time positioning of the mobile camera in a unified coordinate system. The system integrates VGGT for high-precision pose estimation and combines it with MASt3r-SLAM for robust real-time tracking. Alignment between different coordinate systems is achieved through rotation matrix transformation and Procrustes analysis. User evaluation confirms the effectiveness of the system in indoor navigation scenarios and is expected to help visually impaired users. Although the framework is sensitive to motion blur, it shows strong flexibility, high accuracy, and effective use of multi-source image data. Future work will focus on enhancing robustness to blurred inputs, extending the framework to support 4D scene reconstruction, and developing advanced auxiliary navigation functions to pave the way for practical applications.}}, author = {{Zeng, Jiuming and Hua, Yanling}}, language = {{eng}}, note = {{Student Paper}}, title = {{A Joint Sensing Framework for Real-time Mobile Camera Localization}}, year = {{2025}}, }