Accurate Indoor Localization through Constrained Visual SLAM

Olivier Gomez (1, 2) , Achkan Salehi (1, 2) , Vincent Gay-Bellile (1, 2) , Mathieu Carrier (1, 2)

1 LVIC - Laboratoire Vision et Ingénierie des Contenus
2 DIASI - Département Intelligence Ambiante et Systèmes Interactifs

— In this paper, we focus on navigation in indoor environments using Visual SLAM (VSLAM). We propose an approach to suppress the known drifting issue of VSLAM and express its localization in building coordinate frame. It relies on a database built offline through a coarse to fine strategy that registers and refines a VSLAM reconstruction by taking advantage of the building 3D model. The database can then be extended online when the user goes out and comes back in the known environment. We present experimental results on synthetic and real data.

Keywords: SLAM Simultaneous Localisation and Mapping; Indoor navigation; 3D database

Domains: Signal and Image processing

Olivier Gomez, Achkan Salehi, Vincent Gay-Bellile, Mathieu Carrier. Accurate Indoor Localization through Constrained Visual SLAM. 2017 International Conference on Indoor Positioning and Indoor Navigation (IPIN), 2017. ⟨cea-01830464⟩