PEnG: Pose-Enhanced Geo-Localisation
- 27 February 2025
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (4) , 3835-3842
Abstract
Cross-view Geo-localisation is typically performed at a coarse granularity, because densely sampled satellite image patches overlap heavily. This heavy overlap would make disambiguating patches very challenging. However, by opting for sparsely sampled patches, prior work has placed an artificial upper bound on the localisation accuracy that is possible. Even a perfect oracle system cannot achieve accuracy greater than the average separation of the tiles. To solve this limitation, we propose combining cross-view geo-localisation and relative pose estimation to increase precision to a level practical for real-world application. We develop PEnG, a 2-stage system which first predicts the most likely edges from a city-scale graph representation upon which a query image lies. It then performs relative pose estimation within these edges to determine a precise position. PEnG presents the first technique to utilise both viewpoints available within cross-view geo-localisation datasets, referring to this as Multi-View Geo-Localisation (MVGL). This enhances accuracy to a sub-metre level, with some examples achieving centimetre level precision. Our proposed ensemble achieves state-of-the-art accuracy - with relative Top-5 m retrieval improvements on previous works of 213%. Decreasing the median Euclidean distance error by 96.90% from the previous best of 734 m down to 22.77 m, when evaluating with $90^{\circ }$ horizontal FOV images.Keywords
All Related Versions
Funding Information
- Engineering and Physical Sciences Research Council (EP/S035761/1)
- FlexBot - InnovateUK (10067785)
This publication has 34 references indexed in Scilit:
- Grounding Image Matching in 3D with MASt3R2024
- Map-Free Visual Relocalization: Metric Pose Relative to a Single Image2022
- Visual Cross-View Metric Localization with Dense Uncertainty Estimates2022
- DSAC — Differentiable RANSAC for Camera Localization2017
- Localizing and Orienting Street Views Using Overhead Imagery2016
- Learning deep representations for ground-to-aerial geolocalization2015
- Scene Coordinate Regression Forests for Camera Relocalization in RGB-D Images2013
- From structure-from-motion point clouds to fast location recognition2009
- ImageNet: A large-scale hierarchical image database2009
- Monocular Vision for Mobile Robot Localization and Autonomous NavigationInternational Journal of Computer Vision, 2007