Efficacy of Monocular Depth Estimation in Lunar Environments

Michael W. Lanighan; Kenji Brameld; Seth Gee

cs.ROOct 27, 2026

Efficacy of Monocular Depth Estimation in Lunar Environments

Authors: Michael W. Lanighan, Kenji Brameld, Seth Gee

Organizations: TRACLabs Inc., 1331 Gemini Street; Houston, TX 77058.

Abstract

Recent advances in monocular depth estimation have demonstrated highly accurate 3D reconstructions within in-distribution terrestrial domains. This study investigates the efficacy of several state-of-the-art approaches to reconstruct out-of-distribution lunar environments using the NASA Ames Polar Optical Lunar Analog Reconstruction (POLAR) stereo dataset and Apollo 17 surface imagery. By leveraging these datasets, we establish a robust analog-to-field validation pipeline that evaluates quantitative performance against ground-truth physical analogs and qualitative performance against historical field data. Our results indicate that select state-of-the-art techniques can successfully generalize to produce high-fidelity relative 3D reconstructions of the unstructured lunar surface.

Figures & tables

Panel of Figure 1 — Fig. 1 : Input Apollo 17 image (left) and the resulting depth images and point clouds (middle and right columns) generated by the specified monocular depth estimation techniques. Note the consistent reconstruction of the barren, textureless lunar landscape, even in the presence of shadows.

SILog	$1\sigma$ Error Factor ( $\rho$ )	$2\sigma$ Error Factor ( $\rho^{2}$ )
2.5	$1.03\times$	$1.05\times$
5	$1.05\times$	$1.11\times$
10	$1.11\times$	$1.22\times$
50	$1.65\times$	$2.72\times$
100	$2.72\times$	$7.39\times$

TABLE I : Statistical Interpretation of SILog values

Fig. 2 : Reconstructions for Terrain 1, capture position A of the POLAR stereo dataset featuring a rock-strewn terrain. The figure shows the input, the masked ground truth depth, and the masked reconstructions of each approach used to calculate SILog.

Fig. 3 : Reconstructions for Terrain 1, capture position C of the POLAR stereo dataset featuring a rock-strewn terrain with lens flaring. The figure shows the input, the masked ground truth depth, and the masked reconstructions of each approach used to calculate SILog.

Fig. 4 : Reconstructions for Terrain 6, capture position A of the POLAR stereo dataset featuring a barren terrain. The figure shows the input, the masked ground truth depth, and the masked reconstructions of each approach used to calculate SILog.

Fig. 5 : Reconstructions for Terrain 6, capture position C of the POLAR stereo dataset featuring a barren terrain with lens flaring. The figure shows the input, the masked ground truth depth, and the masked reconstructions of each approach used to calculate SILog.

Fig. 6 : Reconstructions for Terrain 11, capture position A of the POLAR stereo dataset featuring a large crater. The figure shows the input, the masked ground truth depth, and the masked reconstructions of each approach used to calculate SILog.

Fig. 7 : Reconstructions for Terrain 11, capture position C of the POLAR stereo dataset featuring a large crater with lens flaring. The figure shows the input, the masked ground truth depth, and the masked reconstructions of each approach used to calculate SILog.

	T01-Rocks		T06-Smooth		T11-FreshCrater
Approach	Pos A	Pos C	Pos A	Pos C	Pos A	Pos C
MoGe-2	1.62	1.68	1.93	2.90	2.56	3.16
OmniVGGT	1.91	2.63	3.19	3.86	3.26	3.63
VGGT	2.92	4.34	3.46	3.01	4.84	4.22
MapAnything	16.10	16.95	2.02	1.28	19.51	21.26
Depth Anything v2	12.48	6.16	16.86	17.21	13.88	12.82

TABLE II : Detailed SILog Results per Terrain and Capture Position

Approach	Mean SILog $\downarrow$	Std Dev	Mean Coverage (%) $\uparrow$
MoGe-2	2.31	0.60	100.00
OmniVGGT	3.08	0.65	100.00
VGGT	3.80	0.71	100.00
MapAnything	12.85	8.10	99.12
Depth Anything v2	13.24	3.65	100.00
Marigold	70.50	47.79	99.27

TABLE III : Overall Performance Metrics by Approach

Panel of Figure 11 — Fig. 8 : Input Apollo 17 image (top) and the resulting depth images (left) and point clouds (right) generated by the specified monocular depth estimation techniques.

Monocular Depth Estimation

Monocular depth estimation infers scene geometry and distance from a single camera image, enabling 3D perception without stereo or depth sensors.

See all 136 papers

LuMon: A Comprehensive Benchmark and Development Suite with Novel Datasets for Lunar Monocular Depth Estimation
Apr 10, 2026Aytaç Sekmen, Fatih Emre Gunes, Furkan Horoz +9Monocular Depth Estimation Metric Depth Estimation
Department of Computer Engineering, Middle East Technical University, Ankara, Turkey · ROMER Middle East Technical University, Ankara, Turkey · Department of Aerospace Engineering, Middle East Technical University, Ankara, Turkey
MoonGS: High-quality Representation of the Lunar Surface via Gaussian Splatting Using Robust Depth Features from Image Pairs
Oct 5, 2026Yun Jiang, Bo Zheng, Yingying Zhang +73D Gaussian Splatting 3D Reconstruction
Jilin University, Changchun, China · Shanghai Aerospace Control Technology Institute, Shanghai, China · Beijing Institute of Control Engineering, Beijing, China +2
LunarDepthNet: Generation of Digital Elevation Models using Deep Learning and Monocular Satellite Images
Apr 22, 2026Aaranay Aadi, Jai Gopal Singla, Amitabh +3U-Net Monocular Depth Estimation
School of Computer Science and Engineering Manipal University Jaipur Jaipur, India · Space Applications Centre (SAC) Indian Space Research Organisation (ISRO) Ahmedabad, India · Department of IoT and Intelligent Systems Manipal University Jaipur Jaipur, India +1

Apr 10, 2026J/K move · Enter open · S save

LuMon: A Comprehensive Benchmark and Development Suite with Novel Datasets for Lunar Monocular Depth Estimation

Aytaç Sekmen, Fatih Emre Gunes, Furkan Horoz, Hüseyin Umut Işık+8

Department of Computer Engineering, Middle East Technical University, Ankara, Turkey · ROMER Middle East Technical University, Ankara, Turkey · Department of Aerospace Engineering, Middle East Technical University, Ankara, Turkey

Monocular Depth Estimation Metric Depth Estimation Synthetic-to-Real Domain Adaptation

Open paper AskPDF arXiv

Monocular Depth Estimation (MDE) is crucial for autonomous lunar rover navigation using electro-optical cameras. However, deploying terrestrial MDE networks to the Moon brings a severe domain gap due to harsh shadows, textureless regolith, and zero atmospheric scattering. Existing evaluations rely on analogs that fail to replicate these conditions and lack actual metric ground truth. To address this, we present LuMon, a comprehensive benchmarking framework to evaluate MDE methods for lunar exploration. We introduce novel datasets featuring high-quality stereo ground truth depth from the real Chang'e-3 mission and the CHERI dark analog dataset. Utilizing this framework, we conduct a systematic zero-shot evaluation of state-of-the-art architectures across synthetic, analog, and real datasets. We rigorously assess performance against mission critical challenges like craters, rocks, extreme shading, and varying depth ranges. Furthermore, we establish a sim-to-real domain adaptation baseline by fine tuning a foundation model on synthetic data. While this adaptation yields drastic in-domain performance gains, it exhibits minimal generalization to authentic lunar imagery, highlighting a persistent cross-domain transfer gap. Our extensive analysis reveals the inherent limitations of current networks and sets a standard foundation to guide future advancements in extraterrestrial perception and domain adaptation.

Abstract

Monocular Depth Estimation

Explore similar work

LuMon: A Comprehensive Benchmark and Development Suite with Novel Datasets for Lunar Monocular Depth Estimation

MoonGS: High-quality Representation of the Lunar Surface via Gaussian Splatting Using Robust Depth Features from Image Pairs

LunarDepthNet: Generation of Digital Elevation Models using Deep Learning and Monocular Satellite Images