A Formally Verified Framework for Indoor Positioning with Wi-Fi Fingerprinting, BoVW, and ORB

Indoor positioning remains a challenging task due to signal variability, repetitive building layouts, and the limitations of relying on a single sensing modality. This paper presents a hybrid indoor localization framework that integrates Wi-Fi fingerprinting, Bag of Visual Words (BoVW) visual recognition, and ORB feature matching to achieve high accuracy using low-cost, widely available sensors. Wi-Fi fingerprinting is first used to estimate a coarse user location by comparing real-time RSSI measurements with a pre-constructed radio map. BoVW is then applied as a fine localization step to identify the most visually similar cell within the reduced search region. When discrepancies arise between Wi-Fi and BoVW predictions, the ORB algorithm performs keypoint-level comparisons to select the most visually consistent location. The proposed system was evaluated across multiple floors in two university buildings (The British University in Egypt, and University of Hertfordshire in Egypt), using a dense grid of reference cells and bidirectional image and RSSI collection. Experimental results demonstrate that the hybrid approach significantly improves localization accuracy compared to using Wi-Fi or visual methods alone, reliably identifying the correct cell or its immediate neighbor even in visually repetitive corridors and daytime conditions with pedestrian presence. By leveraging existing Wi-Fi infrastructure and smartphone cameras, the framework offers a practical, scalable, and cost-effective solution for real-world indoor positioning applications.