Moon in 3D: Streamlined Automated DEM Production from LROC NAC Stereo Pairs

MOON IN 3D: STREAMLINED AUTOMATED DEM PRODUCTION FROM LRO NAC STEREO PAIRS. N. Schmedemann1, C. H. van der Bogert1, H. Hiesinger1, 1Institut für Planetologie, Universität Münster, Wilhelm-Klemm-Strasse 10, 48149 Münster, Germany ([email protected]). Introduction: The Lunar Reconnaissance Orbiter Narrow Angle Camera (LRO NAC) [1] provides high‐resolution images of the lunar surface, enabling detailed geological and geomorphological investigations including surface details critical for mapping impact craters, tectonic features, and regolith properties. Digital Elevation Models (DEMs) derived from NAC stereo image pairs are an indispensable dataset because they offer a 3D perspective on surface morphology facilitating quantitative terrain analysis, slope computation, and volume estimations. These and other products are essential for hazard assessment, landing site selection, and understanding the geological evolution of the Moon. Recently, the LROC team has focused on automating DEM generation to cope with the vast number of stereo pairs (>12,150) now available (Fig. 1). However, due to the complexity and computational demands of stereo processing, only ~2,136 pairs are processed to date. Thus, we developed an automated pipeline using state-of-the-art tools and custom scripts (Fig. 2). Methods: Data Section: The selection of appropriate stereo pairs is based on overlapping conditions and temporal proximity (within 10 LRO orbit periods), ensuring that the images provide sufficient parallax for reliable stereo matching. An initial Matlab script filters and writes these selected pairs to a text file, avoiding duplicate processing and ensuring consistency in dataset selection. Preprocessing: Raw NAC images are calibrated and map-projected using USGS ISIS software (versions 3.5.2 and 5.0.0) [2]. Image sub-sections are mosaicked to generate seamless image pairs, ready for stereo processing. A master script checks the processing queue and coordinates subsequent actions. Main Processing: The stereo matching and DEM generation is done in the Ames Stereo Pipeline (ASP) [3]. The processing script reads projection parameters and image names from parameter files, dynamically adjusts map settings, and executes several USGS ISIS tools for bundle adjustment, mosaicking, and DEM production. This stage utilizes parallel processing to manage computational load, ensuring that tasks such as point cloud alignment and DEM generation are efficiently handled. In addition, reference data from LRO LOLA are used to correct absolute height, roll, and pitch of the generated DEMs. Postprocessing: A postprocessing script organizes and archives the final products (e.g., Fig. 3) by copying map cubes, mosaics, DEMs, orthoimages, and good-pixel maps into a designated folder. Derivative products including slope maps, hillshades, and colored hillshades are also made. Throughout the pipeline, robust error handling and quality control measures are implemented to manage common challenges such as projection mismatches and processing interruptions. This automation drastically reduces manual intervention and shortens the processing cycle to a streamlined operation. Fig. 1: Global map of the Moon with footprints of available stereo image pairs (red) and processed DEMs (yellow) from stereo pairs collected for DEM generation. Fig. 2: Flowchart of the automated DEM generation pipeline. Results: The automated processing pipeline has demonstrated significant promise in the generation of high-quality DEMs from NAC data. Of the ~12,150 stereo pairs available, ~2,140 pairs have been processed into DEM products (Fig. 1). These products are validated against existing LOLA data, showing strong consistency in elevation and morphological features. Detailed assessments reveal that the pipeline not only maintains a high standard of geometric accuracy but also efficiently handles data throughput, which is critical given the volume of data. The workflow is designed to be scalable, allowing for the integration of additional processing modules and quality control measures as needed. Ongoing enhancements will include photometric corrections, iterative DEM processing and fine-tuning smoothing parameters, as well as a final shape from shading stage to further improve DEM quality and resolution. Moreover, the creation of footprint shapefiles for processed DEMs enables efficient archiving and retrieval, supporting broader dissemination through web interfaces and PDS-standard data products. Future work will focus on addressing residual processing challenges, optimizing the pipeline for even larger datasets, and further integration of ancillary data to refine the DEM products. References: [1] Robinson, M. S., et al. (2010). Space Sci. Rev., 150(1), 81–124, https://doi.org/ 10.1007/s11214-010-9634-2. [2] Becker, K.J., et al. (2013). Lunar and Planetary Science Conference, p. 2829. [3] Beyer, Ross A., el al. (2018). Earth and Space Science, 5 https://doi.org/10.1029/2018EA000 409. Fig. 3: Example for ASP produced datasets for part of Rima Hyginus using one NAC stereo image pair. From left to right: LRO -NAC mosaic, DEM, ortho image, goodpixel map, hillshade and colored hillshade.