A Visual-inertial Localization Algorithm using Opportunistic Visual Beacons and Dead-Reckoning for GNSS-Denied Large-scale Applications

TL;DR

This work tackles continuous pedestrian localization in GNSS-denied urban environments by fusing visual beacon signals with inertial navigation. It introduces MSGC-NetVLAD, a lightweight visual place recognition network built from multi-scale grouped convolutions and NetVLAD pooling, and couples it with a magnetic disturbance–robust PDR (MDR-PDR). A Kalman filter with gross error suppression (GES) integrates VPR corrections with PDR drift, applying a reliability threshold to mitigate erroneous visual observations. Experimental results on public and private datasets show that MSGC-NetVLAD achieves competitive Recall@1 with far fewer parameters than VGG16-based methods, while the PDR-VPR fusion significantly improves localization accuracy and continuity in GNSS-denied settings, especially in dense urban features. The approach offers a practical, low-cost path toward AR-enabled, large-scale localization on mobile devices.

Abstract

With the development of smart cities, the demand for continuous pedestrian navigation in large-scale urban environments has significantly increased. While global navigation satellite systems (GNSS) provide low-cost and reliable positioning services, they are often hindered in complex urban canyon environments. Thus, exploring opportunistic signals for positioning in urban areas has become a key solution. Augmented reality (AR) allows pedestrians to acquire real-time visual information. Accordingly, we propose a low-cost visual-inertial positioning solution. This method comprises a lightweight multi-scale group convolution (MSGC)-based visual place recognition (VPR) neural network, a pedestrian dead reckoning (PDR) algorithm, and a visual/inertial fusion approach based on a Kalman filter with gross error suppression. The VPR serves as a conditional observation to the Kalman filter, effectively correcting the errors accumulated through the PDR method. This enables the entire algorithm to ensure the reliability of long-term positioning in GNSS-denied areas. Extensive experimental results demonstrate that our method maintains stable positioning during large-scale movements. Compared to the lightweight MobileNetV3-based VPR method, our proposed VPR solution improves Recall@1 by at least 3\% on two public datasets while reducing the number of parameters by 63.37\%. It also achieves performance that is comparable to the VGG16-based method. The VPR-PDR algorithm improves localization accuracy by more than 40\% compared to the original PDR.

Figures (9)

• Figure 1: Overview of the proposed PDR-VPR Algorithm
• Figure 2: Overview of MSGC-NetVLAD
• Figure 3: Optional Structure for MSGC-Block
• Figure 4: Trajectory comparison of Trajectory 1 in a sparse feature environment. The red points are correctly recognized, while the green ones are unrecognized. The indices of the green points from bottom to top are 1, 3, and 4. PDR-VPR w/ GES refers to the proposed visual-inertial localization algorithm; MDR-PDR represents the proposed dead-reckoning method.
• Figure 5: Trajectory comparison of Trajectory 2 in dense feature environment. The red points are correctly recognized, while the green ones are unrecognized. The index of the green point is 1. PDR-VPR w/ GES refers to the proposed visual-inertial localization algorithm; MDR-PDR represents the proposed dead-reckoning method.