Sub-metre Lunar DEM Generation and Validation from Chandrayaan-2 OHRC Multi-View Imagery Using an Open-Source Pipeline

Aaranay Aadi; Jai Singla; Nitant Dube; Oleg Alexandrov

cs.CVApr 1, 2026

Sub-metre Lunar DEM Generation and Validation from Chandrayaan-2 OHRC Multi-View Imagery Using an Open-Source Pipeline

Authors: Aaranay Aadi, Jai Singla, Nitant Dube, Oleg Alexandrov

Organizations: School of Computer Science and Engineering Manipal University Jaipur Jaipur, India · Space Applications Centre (SAC) Indian Space Research Organisation (ISRO) Ahmedabad, India · Intelligent Robotics Group NASA Ames Research Center Moffet Field, CA, USA

Abstract

High-resolution digital elevation models (DEMs) of the lunar surface are essential for surface mobility planning, landing site characterization, and planetary science. The Orbiter High Resolution Camera (OHRC) on board Chandrayaan-2 has the best ground sampling capabilities of any lunar orbital imaging currently in use by acquiring panchromatic imagery at a resolution of roughly 20-30 cm per pixel. This work presents, for the first time, the generation of sub-metre DEMs from OHRC multi-view imagery using an exclusively open-source pipeline. Candidate stereo pairs are identified from non-paired OHRC archives through geometric analysis of image metadata, employing baseline-to-height (B/H) ratio computation and convergence angle estimation. Dense stereo correspondence and ray triangulation are then applied to generate point clouds, which are gridded into DEMs at effective spatial resolutions between approximately 24 and 54 cm across five geographically distributed lunar sites. Absolute elevation consistency is established through Iterative Closest Point (ICP) alignment against Lunar Reconnaissance Orbiter Narrow Angle Camera (NAC) Digital Terrain Models, followed by constant-bias offset correction. Validation against NAC reference terrain yields a vertical RMSE of 5.85 m (at native OHRC resolution), and a horizontal accuracy of within a pixel (approximately 30 cm) assessed by planimetric feature matching.

Figures & tables

Region	Approx. Location	Product ID (Image 1)	Product ID (Image 2)	GSD (cm/px)	Alt 1 (km)	Alt 2 (km)
1	$\sim\!69^{\circ}$ S, $32^{\circ}$ E	20240425T1209509264	20240425T1406019344	0.26	101.90	102.35
2	$\sim\!5^{\circ}$ N, $234^{\circ}$ E	20220914T1033119094	20220914T0835371412	0.32	124.26	123.97
3	$\sim\!69^{\circ}$ S, $342^{\circ}$ E	20230303T0350447888	20230303T0152168201	0.25	99.52	99.43
4	$\sim\!13^{\circ}$ S, $25^{\circ}$ E	20210402T0155096873	20210401T2200364910	0.26	103.81	104.64
5	$\sim\!4^{\circ}$ N, $230^{\circ}$ E	20220321T0525226030	20220321T0326369085	0.19	76.66	76.32

Table 1: OHRC image dataset: acquisition locations, product IDs, native ground sampling distances, and spacecraft altitudes. All Product IDs share the common prefix ch2_ohr_nrp_ .

Figure 1: Overview of the OHRC DEM generation pipeline. Stages proceed from multi-view OHRC imagery and stereo pair selection through camera geometry initialisation, sensor model generation (CSM), bundle adjustment (camera parameter refinement), dense stereo correspondence and feature matching, triangulated 3D point cloud reconstruction, DEM interpolation and surface reconstruction, and finally DEM co-registration and void filling to yield the final terrain product.

Region	Approx. Location	B/H Ratio	Conv. Angle ( ∘ )	DEM Resolution
1	$\sim\!69^{\circ}$ S, $32^{\circ}$ E	0.396	$\sim\!22^{\circ}$	31 cm
2	$\sim\!5^{\circ}$ N, $234^{\circ}$ E	0.414	$\sim\!23^{\circ}$	54 cm
3	$\sim\!69^{\circ}$ S, $342^{\circ}$ E	0.559	$\sim\!31^{\circ}$	30 cm
4	$\sim\!13^{\circ}$ S, $25^{\circ}$ E	0.877	$\sim\!47^{\circ}$	48 cm
5	$\sim\!4^{\circ}$ N, $230^{\circ}$ E	1.161 ‡	$\sim\!61^{\circ}$ †	24 cm

Table 2: Stereo geometry and generated DEM characteristics for each region, ordered by increasing B/H ratio.

Figure 2: DEMs for four of the reconstructed regions. Region 5 is demonstrated in section 7 for its high void fraction. Regions 2 and 4 have well-filled DEMs with minimal voids. Region 1 was not fully filled using NAC due to the presence of holes in the corresponding NAC DTM. Region 3 has no corresponding NAC reference available.

Panel of Figure 3 — Figure 3: Visual feature comparison of OHRC generated DEM vs NAC DTM (both hillshaded) for Sites 1 and 2, randomly selected from Region 4.

Panel of Figure 4 — Figure 4: Visual feature comparison of OHRC generated DEM vs NAC DTM (both hillshaded) for Sites 3 and 4, randomly selected from Region 2.

Region	RMSE (cm)
1	298.96
2	850.42
3	— (no NAC DTM available)
4	606.66
5	— (excluded; high void fraction)
Mean (Regions 1, 2, 4)	585.35 $\approx$ 5.85 m

Table 3: Vertical RMSE by region (computed from the standard deviation of elevation residuals after bias correction, compared against the NAC DTM).

Figure 5: Terrain profile comparison across Region 4. Horizontal axis: distance along profile transect (m). Vertical axis: elevation (m). Red curve: OHRC DEM. Green curve: NAC DTM reference.

Figure 6: Terrain profile comparison across Region 1. Horizontal axis: distance along profile transect (m). Vertical axis: elevation (m). Red curve: OHRC DEM. Green curve: NAC DTM reference.

Figure 7: Terrain profile comparison across Region 2. Horizontal axis: distance along profile transect (m). Vertical axis: elevation (m). Red curve: OHRC DEM. Green curve: NAC DTM reference.

Figure 8: Terrain profile comparison across a small patch of Region 2. Horizontal axis: distance along profile transect (m). Vertical axis: elevation (m). Red curve: OHRC DEM. Green curve: NAC DTM reference.

Pipeline Stage	Runtime (per pair)
Data ingestion & SPICE init	$\sim$ 1–2 min
Sensor model generation	$<$ 1 min
Bundle adjustment	$\sim$ 10–20 min
Dense stereo correspondence	$\sim$ 10–30 min
Point cloud rasterisation	$\sim$ 30 min – 2 h
ICP alignment & offset corr.	$\sim$ 5–15 min

Table 4: Approximate wall-clock runtimes per stereo pair for each pipeline stage. Hardware: Intel Xeon Gold 6530

\times

128 cores, 1 TB RAM.

Figure 9: Region 5 demonstrated with high void fraction

Figure 10: Region-of-interest stereo reconstruction. Left and centre: the left-camera and right-camera cropped image chips in native pixel coordinates. Right: the reconstructed DEM for the same region.

Remote Sensing

Remote sensing uses satellite, aerial, and ground-based observations to measure and map Earth and other planetary surfaces.

See all 528 papers

DEM Refinement and Validation on the Lunar Surface Using Shape-from-Shading with Chandrayaan-2 OHRC Imagery
Apr 19, 2026Aaranay Aadi, Jai Gopal Singla, Nitant DubeRemote Sensing
School of Computer Science and Engineering, Manipal University Jaipur Jaipur, India · Space Applications Centre (SAC), Indian Space Research Organisation (ISRO) Ahmedabad, India
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
Towards Seamless Lunar Mosaics: Deep Radiometric Normalization for Cross-Sensor Orbital Imagery Using Chandrayaan-2 TMC Data
Apr 28, 2026Pratincha Singh, Jai Gopal Singla, Prashant Hemrajani +3Remote Sensing Image Understanding Remote Sensing
Department of IOT and Intelligent Systems Manipal University Jaipur Jaipur,India · Space Applications Centre Indian Space Research Organization Ahmedabad,India

Apr 19, 2026J/K move · Enter open · S save

DEM Refinement and Validation on the Lunar Surface Using Shape-from-Shading with Chandrayaan-2 OHRC Imagery

Aaranay Aadi, Jai Gopal Singla, Nitant Dube

School of Computer Science and Engineering, Manipal University Jaipur Jaipur, India · Space Applications Centre (SAC), Indian Space Research Organisation (ISRO) Ahmedabad, India

Remote Sensing

Open paper AskPDF arXiv

This study presents a Shape from Shading (SfS) framework to enhance sub-metre resolution lunar digital elevation models (DEMs) using imagery from the Orbiter High Resolution Camera (OHRC) aboard Chandrayaan-2. The framework applies SfS to an independent OHRC image of the same region, enabling SfS not just as a refinement tool, but as a source of new topographic data, unconstrained by stereo baseline limitations. The method is applied across three lunar sites, including the Cyrillus crater, the Vikram landing region, and the lunar south pole (Mons Mouton), with a systematic three-stage parameter sweep on the SfS smoothness weight. Results show measurable topographic enhancement, particularly in surface slope statistics, revealing fine-scale crater morphology previously unresolved. A limiting case is also characterized, where large pitch angle separation between the shading image and stereo pair reduces SfS sensitivity, and partial footprint coverage of the shading image is identified as a factor influencing spatially variable enhancement quality.

Abstract

Remote Sensing

Explore similar work

DEM Refinement and Validation on the Lunar Surface Using Shape-from-Shading with Chandrayaan-2 OHRC Imagery

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

Towards Seamless Lunar Mosaics: Deep Radiometric Normalization for Cross-Sensor Orbital Imagery Using Chandrayaan-2 TMC Data