June 6, 2025, Friday, 03:00PM (GMT+03.00, Moscow), IKI, Room 200

Maxim Pupkov
(IKI, Saratov State University)

Reconstruction of the lunar surface using photogrammetry and neural networks methods

( M. Pupkov1, A. Prosvetov1, V. Marchuk2, A. Govorov1, O. Yushkova2, A. Andreev1, V. Nazarov1)
1 - Space Research Institute of the Russian Academy of Sciences (IKI)
2 - Kotelnikov Institute of Radioengineering and Electronics

Abstract:

Thorough and precise lunar surface modeling forms the foundation for effective mission design and successful execution of lunar exploration initiatives. Comprehensive digital reconstructions are critical when identifying suitable sites for landers as well as plotting traverses for robotic vehicles. Historically, the use of photogrammetry has underpinned much of the terrain reconstruction work. In recent times, however, advanced machine learning algorithms – including those based on neural networks, such as Neural Radiance Fields and Gaussian Splatting – have brought significant improvements. These technologies enable more nuanced modeling, allowing better handling of challenging lighting and textural variations.

Furthermore, contemporary lunar missions routinely collect ancillary data, such as surface temperature profiles and radarograms across multiple frequency bands. Integration of this additional information into the neural network learning structure can significantly enhance the accuracy and robustness of surface reconstructions, offering a more comprehensive representation of lunar terrain characteristics.

The aim of the current study is to apply modern photogrammetric and new reconstruction methods to analyze images of the Moon's surface that differ significantly from conventional images: they do not contain penumbra, translucent objects, and objects with glare. The paper compares the reconstruction of the surface by three different methods – classical photogrammetry, Neural Radiance Fields and Gaussian Splatting, and performs a qualitative analysis of the reconstruction on images taken at different lighting angles corresponding to different times of the satellite's passage over the Moon's surface.




Video recording of the report is provided upon request for personal use only. The request for access can be sent to the organizing committee of the seminar by e-mail to mci@cosmos.ru. The materials provided on request have no publication status and are not intended for mass distribution.