Indoor localization remains a challenging problem due to the limitations of Global Navigation Satellite Systems (GNSS) in enclosed environments and the practical constraints of infrastructure-based solutions such as Bluetooth, WiFi, Ultra Wide Band (UWB), and Radio-Frequency (RF) systems. Recent vision-based approaches have shown promising results, but they often require manually annotated pose data, computationally expensive reconstruction pipelines, or additional sensor calibration and alignment procedures. In this paper, we propose a lightweight vision-based framework for indoor visitor localization in museum environments using frames extracted from smartphone-recorded videos. We approach localization as a content-based image retrieval task, matching query frames against a reference database generated from 360-degree LiDAR acquisitions captured with a Matterport scanning system. By directly leveraging the spatial metadata and acquisition graph natively provided by Matterport scans, the framework automatically associates visual observations with spatial coordinates without requiring any manual annotation. We evaluate the proposed framework at the Museum of Archaeology of the University of Catania (MAUC), demonstrating strong localization performance in a real-world cultural heritage environment. Experimental results show that the proposed approach achieves up to 88.54% exact viewpoint and 91.56% tolerance-aware localization accuracy, while operating in real time at only 80 ms per frame on lightweight hardware.

Where Am I in the Museum? Metadata-Driven Visual Localization via Image Retrieval from 360-Degree LiDAR Scans

Allegra D.;Gueli A. M.;Stanco F.
2026-01-01

Abstract

Indoor localization remains a challenging problem due to the limitations of Global Navigation Satellite Systems (GNSS) in enclosed environments and the practical constraints of infrastructure-based solutions such as Bluetooth, WiFi, Ultra Wide Band (UWB), and Radio-Frequency (RF) systems. Recent vision-based approaches have shown promising results, but they often require manually annotated pose data, computationally expensive reconstruction pipelines, or additional sensor calibration and alignment procedures. In this paper, we propose a lightweight vision-based framework for indoor visitor localization in museum environments using frames extracted from smartphone-recorded videos. We approach localization as a content-based image retrieval task, matching query frames against a reference database generated from 360-degree LiDAR acquisitions captured with a Matterport scanning system. By directly leveraging the spatial metadata and acquisition graph natively provided by Matterport scans, the framework automatically associates visual observations with spatial coordinates without requiring any manual annotation. We evaluate the proposed framework at the Museum of Archaeology of the University of Catania (MAUC), demonstrating strong localization performance in a real-world cultural heritage environment. Experimental results show that the proposed approach achieves up to 88.54% exact viewpoint and 91.56% tolerance-aware localization accuracy, while operating in real time at only 80 ms per frame on lightweight hardware.
2026
Content-Based Image Retrieval
Indoor Museum Localization
LiDAR-based visual database
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11769/738710
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact