AeroMap3D: Anchoring Monocular UAV 6-DoF Localization to Visual-Geometric-Semantic Map Priors

2026-07-15

AI Summary

This work addresses the challenges of monocular visual localization for UAVs in GNSS-denied environments, where cross-view discrepancies between aerial imagery and satellite maps, along with inconsistencies between digital elevation models (DEMs) and actual building structures, introduce significant errors. To mitigate these issues, the authors propose a lightweight, fine-tuning-free adapter that corrects scale and yaw biases to align onboard images with a visual-geometric-semantic map prior. Reliable feature matches are selected by integrating OpenStreetMap semantic information, and dense matching is refined over the DEM. Six-degree-of-freedom continuous localization is achieved through a combination of RANSAC-PnP and a delayed-state extended Kalman filter. Evaluated across eight test sites in Austin over 55 km of flight, the method achieves an average 3D localization error of only 5.88 meters, with all trajectory errors remaining below 50 meters.

Abstract

We present AeroMap3D, a monocular 6-DoF UAV localization system that anchors onboard imagery to visual, geometric, and semantic map priors for GNSS-denied navigation. AeroMap3D addresses two fundamental challenges in map-referenced aerial localization: the cross-view discrepancy between UAV imagery and satellite maps, and the structural inconsistency between bare-earth digital elevation models (DEMs) and urban scenes. First, we introduce a lightweight adapter that enables a dense matcher pretrained on internet-scale generic data to perform reliable UAV-to-map registration without finetuning. By estimating the scale ratio and yaw offset between the UAV image and map tile, the adapter removes the dominant geometric misalignment induced by altitude, camera field of view, and heading before dense correspondence estimation. Second, AeroMap3D lifts 2D UAV-map correspondences onto DEM terrain while using OpenStreetMap annotations to reject semantically unreliable matches before RANSAC-PnP pose estimation, thereby reducing errors caused by unmodeled building heights and off-nadir structures. Delayed map-based pose measurements are further fused with relative-motion priors using a delayed-state EKF for continuous trajectory estimation. Without UAV-Terra3D retraining or tuning, AeroMap3D localizes all trajectories across eight Austin sites within 50 m and achieves 5.88 m mean 3D error over 55 km of flight.

Problem

• UAV localization
• cross-view discrepancy
• structural inconsistency
• GNSS-denied navigation
• 6-DoF pose estimation

Innovation

• monocular localization
• cross-view matching
• visual-geometric-semantic fusion
• delayed-state EKF
• UAV navigation