Reproduce “GNSS-denied geolocalization of UAVs by visual matching of onboard camera images with orthophotos” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jouko Kinnari, Francesco Verdoja, Ville Kyrki
A system for GNSS-denied UAV geolocalization was developed, enabling an arbitrarily oriented camera to match its orthorectified images against 2D orthophoto maps. This approach achieved localization errors below 50m Root-Mean-Square (RMS) after convergence, outperforming odometry-only methods.
Key insights
- Utilizing orthorectification of images from an arbitrarily oriented camera allows for robust geolocalization without a dedicated downward-facing camera, expanding UAV operational flexibility.
- The Monte-Carlo Localization approach can achieve global position convergence even with substantial initial pose uncertainty (e.g., a 200x200m area with unknown heading) after sufficient travel.
- Classical area-based image matching metrics, specifically Moravec correlation, proved effective in discriminating between true and false map matches for georeferencing, even with modest map resolutions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.