Application of visual localization system for visitor guidance

Abstract

This project addresses the issue faced by indoor navigation in GPS-denied environments such as museums due to the reliance on costly infrastructure-dependent solutions like Wi-Fi positioning and Bluetooth beacons. This project presents a real-time indoor localisation system leveraging ORB-SLAM2 for visual feature-based mapping, Supabase for cloud storage and a React-based frontend for 2D/3D visualisation. The system tracks user movement via a monocular camera on an Odroid N2, processes the video feed in real-time and synchronises mapping data to the cloud for dynamic reading. System performance was evaluated based on latency, bandwidth usage and computational load, with results demonstrating efficient real-time tracking and data synchronisation. The system successfully provides infrastructure-free navigation and congestion mapping, enhancing visitor experience in exhibition spaces. Finally, future improvements are also covered which includes optimising multi-user mapping, real-time crowdaware routing and scalability for larger environments.