Metadata-free Georegistration of Ground and Airborne Imagery

TL;DR

This work addresses the problem of georegistering heterogeneous ground and airborne imagery without relying on sensor metadata. It introduces a metadata-free pipeline that uses $NeRF$-based novel-view rendering to bridge domain gaps, combined with the $RoMa$ dense matcher and per-pixel depth to lift 2D correspondences into $3D$ and estimate a similarity transform. The method aligns airborne data to a satellite DSM first, then registers ground data into the airborne frame, demonstrating robust cross-domain performance across three WRIVA sites with quantitative error metrics. By removing dependence on metadata, the approach enables georeferencing of legacy and diverse datasets, enabling integrated multi-view 3D modeling and realistic novel-view rendering across domains.

Abstract

Heterogeneous collections of ground and airborne imagery can readily be used to create high-quality 3D models and novel viewpoint renderings of the observed scene. Standard photogrammetry pipelines generate models in arbitrary coordinate systems, which is problematic for applications that require georegistered models. Even for applications that do not require georegistered models, georegistration is useful as a mechanism for aligning multiple disconnected models generated from non-overlapping data. The proposed method leverages satellite imagery, an associated digital surface model (DSM), and the novel view generation capabilities of modern 3D modeling techniques (e.g. neural radiance fields) to provide a robust method for georegistering airborne imagery, and a related technique for registering ground-based imagery to models created from airborne imagery. Experiments demonstrate successful georegistration of airborne and ground-based photogrammetric models across a variety of distinct sites. The proposed method does not require use of any metadata other than a satellite-based reference product and therefore has general applicability.

Figures (7)

• Figure 1: The proposed georegistration method uses synthetic NeRF renderings generated from airborne imagery to bridge the gap between ground-level and satellite imagery. NeRF rendering allows the control necessary to produce images suitable for 2D matching to both satellite and ground-level imagery, allowing the transfer of satellite geopositioning to ground level images without metadata. Satellite imagery (C) 2025 Maxar
• Figure 2: The 3D airborne model is used to render an nadir-view orthographic image (top) and a series of oblique renderings at regular azimuth spacings. Per-pixel 3D point estimates (right) and uncertainty (not shown) are rendered for each image.
• Figure 3: Dense matching with cyclic consistency results in high-quality reliable matches between the rendered airborne models and satellite imagery. Satellite imagery (C) 2025 Maxar
• Figure 4: Overlapping tiles are extracted from the oblique airborne renderings for matching with ground imagery.
• Figure 5: A consensus among the best matching tiles is taken, so a few incorrect ones do not influence the final transform.