SCAR: Satellite Imagery-Based Calibration for Aerial Recordings

Henry Hölzemann; Michael Schleiss

cs.CVFeb 18, 2026

SCAR: Satellite Imagery-Based Calibration for Aerial Recordings

Authors: Henry Hölzemann, Michael Schleiss

Organizations: Department of Sensor Data and Information Fusion, Fraunhofer FKIE, 53343 Wachtberg, Germany · Chair of Astronautics, University of the Bundeswehr Munich, 85579 Neubiberg, Germany

Abstract

We introduce SCAR, a method for long-term auto-calibration refinement of aerial visual-inertial systems that exploits georeferenced satellite imagery as a persistent global reference. SCAR estimates both intrinsic and extrinsic parameters by aligning aerial images with 2D--3D correspondences derived from publicly available orthophotos and elevation models. In contrast to existing approaches that rely on dedicated calibration maneuvers or manually surveyed ground control points, our method leverages external geospatial data to detect and correct calibration degradation under field deployment conditions. We evaluate our approach on six large-scale aerial campaigns conducted over two years under diverse seasonal and environmental conditions. Across all sequences, SCAR consistently outperforms established baselines (Kalibr, COLMAP, VINS-Mono), reducing median reprojection error by a large margin, and translating these calibration gains into substantially lower visual localization rotation errors and higher pose accuracy. These results demonstrate that SCAR provides accurate, robust, and reproducible calibration over long-term aerial operations without the need for manual intervention.

Figures & tables

Fig. 1: Overview of the proposed SCAR framework for long-term INS–camera calibration refinement. (a) Aerial imagery is matched against satellite or orthophoto data to generate georeferenced anchors, while feature correspondences are tracked across frames to provide multi-view constraints for optimization. (b) Initial intrinsic and extrinsic calibrations are used to evaluate reprojection errors of these anchors. (c) Both absolute anchors and absolute pose measurements (GNSS/INS) are fused in a non-linear factor graph optimization, yielding refined intrinsic and extrinsic calibration parameters.

Fig. 2: Factor graph structure of SCAR. Camera poses

C_{t}

, anchors

X_{j}

, and intrinsics

K

are modeled as variables, constrained by pose priors

\hat{P}^{W}_{\text{CAM},t}

, anchor priors

\hat{X}^{W}_{j}

, calibration prior

K_{\text{init}}

and reprojection factors.

Panel of Figure 3 — Fig. 3: Selected segments of both routes A (orange) and B (blue). For each route, segment 1 is used for calibration, segment 2 for validation.

ID	Flight Time	#GCPs	#Obs.	#Obs./Frame
ID	Flight Time	#GCPs	#Obs.	Mean	Std.	Min.
Training (1.5k frames each; route/segment A1, B1)
A1.i	2022-02-11 11h	23,798	445,662	296.9	58.1	81
A1.ii	2022-02-11 15h	17,265	368,346	245.4	66.8	16
A1.iii	2022-06-10 10h	14,946	445,372	296.7	66.3	56
B1.iv	2022-06-14 10h	14,220	463,043	308.5	43.9	74

TABLE I: Overview of calibration datasets used in the experiments.

ID	Calib.	Pose error		Pose acc. [%] @ m/°
ID	Calib.	Rot. [°]	Trans. [m]	2/2	5/5	10/10
A2.i (Vl)	Kalibr	$2.48\pm 0.12$	$2.19\pm 0.85$	0	90	99
A2.i (Vl)	SCAR	$0.32\pm 0.13$	$1.97\pm 0.79$	51	93	99
A2.ii (Vl)	Kalibr	$2.64\pm 0.22$	$7.23\pm 2.35$	0	23	75
A2.ii (Vl)	SCAR	$0.63\pm 0.25$	$4.75\pm 2.20$	14	53	85
A2.iii (Vl)	Kalibr	$2.25\pm 0.18$	$5.53\pm 1.80$	1	41	84

TABLE II: VL-based evaluation on validation sequences.

ID	Kalibr (ref.)		COLMAP		VINS-Mono ${}^{(\text{a})}$		SCAR [ ours ]
ID	2022-10-04		( $K^{\star}$ ; $T^{\dagger}$ )		( $T^{\star}$ ; $K^{\dagger}$ )		( $K^{\star}$ ; $T^{\dagger}$ )		( $T^{\star}$ ; $K^{\dagger}$ )		( $K^{\star}$ , $T^{\star}$ )
A1.i (Tr)	$45.90$ $\pm$	$7.21$	$45.91$ $\pm$	$7.23$	$13.14$ $\pm$	$4.66$	$46.22$ $\pm$	$7.19$	$5.61$ $\pm$	$2.48$	5.51 $\pm$	$2.44$
A2.i (Val)	$47.16$ $\pm$	$7.50$	$47.14$ $\pm$	$7.52$	$12.76$ $\pm$	$4.39$	$47.44$ $\pm$	$7.52$	$7.19$ $\pm$	$2.83$	7.05 $\pm$	$2.81$
A1.ii (Tr)	$47.45$ $\pm$	$8.54$	$47.90$ $\pm$	$8.58$	$18.52$ $\pm$	$6.39$	$47.07$ $\pm$	$8.64$	$9.06$ $\pm$	$3.94$	8.77 $\pm$	$3.87$
A2.ii (Val)	$47.47$ $\pm$	$8.77$	$47.84$ $\pm$	$8.85$	$17.85$ $\pm$	$6.43$	$46.87$ $\pm$	$8.92$	$10.26$ $\pm$	$4.22$	9.36 $\pm$	$4.05$
A1.iii (Tr)	$46.19$ $\pm$	$7.26$	$46.32$ $\pm$	$7.28$	$30.28$ $\pm$	$6.03$	$46.61$ $\pm$	$7.31$	$7.43$ $\pm$	$3.24$	7.31 $\pm$	$3.20$

TABLE III: Comparison of calibration baselines and scar on median reprojection error across calibration sequences.

Panel of Figure 7 — Fig. 4: Reprojection examples showing reduced errors with SCAR (b,d) compared to Kalibr (a,c), with high (a) and low (c) initial error. The colormap in (a,b) ranges from 0 to 100 px and from 0 to 50 px in (c,d).

Fig. 5: Ablation study of SCAR. Shown is the change when removing nadir assumptions (“no nadir”), merging camera pose optimization into one step (“cam opt joint”), or skipping the final refinement (“no fine adjust”).

Fig. 6: Robustness evaluation on training data size used for SCAR. Plotted are median reprojection errors on validation segments when calibration is estimated from different numbers of training frames.

Camera Calibration

Camera calibration estimates the geometric and optical parameters that connect image measurements to real-world positions and other sensors.

See all 57 papers

Feb 16, 2026

Wrivinder: Towards Spatial Intelligence for Geo-locating Ground Images onto Satellite Imagery

Aligning ground-level imagery with geo-registered satellite maps is crucial for mapping, navigation, and situational awareness, yet remains challenging under large viewpoint gaps or when GPS is unreliable. We introduce Wrivinder, a zero-shot, geometry-driven framework that aggregates multiple ground photographs to reconstruct a consistent 3D scene and align it with overhead satellite imagery. Wrivinder combines SfM reconstruction, 3D Gaussian Splatting, semantic grounding, and monocular depth--based metric cues to produce a stable zenith-view rendering that can be directly matched to satellite context for metrically accurate camera geo-localization. To support systematic evaluation of this task, which lacks suitable benchmarks, we also release MC-Sat, a curated dataset linking multi-view ground imagery with geo-registered satellite tiles across diverse outdoor environments. Together, Wrivinder and MC-Sat provide a first comprehensive baseline and testbed for studying geometry-centered cross-view alignment without paired supervision. In zero-shot experiments, Wrivinder achieves sub-30,m geolocation accuracy across both dense and large-area scenes, highlighting the promise of geometry-based aggregation for robust ground-to-satellite localization.

Chandrakanth Gudavalli, Tajuddin Manhar Mohammed, Abhay Yadav +6

Mayachitra, Inc. · Johns Hopkins University

Cross-View Geo-Localization

Ask PDF

Jun 22, 2026

Scene-agnostic ALS boresight self-calibration

ALS boresight calibration has relied for two decades on dedicated flight patterns over structured scenes containing planar surfaces of varied aspect and slope. While reliable, this approach imposes constraints on the scene content and operations, which limits its applicability to boresight recovery within routine mapping missions. We present a practical approach that substantially relaxes these requirements by replacing plane-based constraints with scene-agnostic point-to-point correspondences extracted automatically from overlapping ALS strips. Two complementary formulations are proposed to estimate boresight with laser vector observations: (i) a simpler parametric adjustment utilizing INS/GNSS trajectory; (ii) a rigorous formulation treating GNSS and raw inertial data within an existing factor-graph, i.e. a dynamic network, where boresight is added as an additional parameter. Both formulations are evaluated across four operational ALS flights equipped with five inertial systems, covering a wide range of flight altitudes, overlap geometries, terrain types and inertial sensor classes. The analysis draws a clear boundary between the legacy plane-based conditioning that falls short outside the calibration scenario and the proposed formulations, which either recover or absorb boresight effects under conventional mapping geometry. Among them, the lightweight formulation is sufficient for boresight recovery using tactical and navigation grade inertial sensors, while the general factor-graph approach is clearly superior when the inertial sensor errors are less observable within an optimal smoother. This supports the hypothesis that, for INS/GNSS trajectory of sufficient quality, the boresight calibration can be performed without particular scene prerequisites during routine mapping operations using a minimum of 3-4 overlapping strips, with either proposed formulation...

Aurélien Brun, Jan Skaloud

aEPFL ENAC ESO, Bâtiment GC - Station 18, 1015 Lausanne, Switzerland

Airborne Laser Scanning

Ask PDF

May 1, 2026

Paired-CSLiDAR: Height-Stratified Registration for Cross-Source Aerial-Ground LiDAR Pose Refinement

We introduce Paired-CSLiDAR (CSLiDAR), a cross-source aerial-ground LiDAR benchmark for single-scan pose refinement: refining a ground-scan pose within a 50 m-radius aerial crop. The benchmark contains 12,683 ground-aerial pairs across 6 evaluation sites and per-scan reference 6-DoF alignments for sub-meter root-mean-square error (RMSE) evaluation. Because aerial scans capture rooftops and canopy while ground scans capture facades and under-canopy, the two modalities share only a fraction of their geometry, primarily the terrain surface, causing standard registration methods and learned correspondence models to converge to metrically incorrect local minima. We propose Residual-Guided Stratified Registration (RGSR), a training-free, geometry-only refinement pipeline that exploits the shared ground plane through height-stratified ICP, reversed registration directions, and confidence-gated accept-if-better selection. RGSR achieves 86.0% [email protected] m and 99.8% [email protected] m on the primary benchmark of 9,012 scans, outperforming both the confidence-gated cascade at 83.7% and GeoTransformer at 76.3%. We validate RMSE-based pose selection with independent survey control and trajectory consistency, and show that added Fourier-Mellin BEV proposals can reduce RMSE while increasing actual pose error under extreme partial overlap. The dataset and code are being prepared for public release.

Montana Hoover, Jing Liang, Tianrui Guan +1

University of Maryland · 2Stanford University

Point Cloud Registration

Ask PDF

Abstract

Camera Calibration

Explore similar work

Wrivinder: Towards Spatial Intelligence for Geo-locating Ground Images onto Satellite Imagery

Scene-agnostic ALS boresight self-calibration

Paired-CSLiDAR: Height-Stratified Registration for Cross-Source Aerial-Ground LiDAR Pose Refinement