Multi-Input CNN for Vision-Based In Situ Analysis of Extraterrestrial Surface Composition

Downloads

Authors

Abstract

Earth’s natural resources are finite, which is why engineers and scientists are increasingly turning their attention to the extraction of materials from celestial bodies. However, beyond the extraction itself, a major challenge remains the localization of valuable substances and the assessment of their quality in situ. In this article, the authors present a flexible and robust method for estimating the content of selected components in heterogeneous mixtures based on the processing of RGB images. The proposed deep learning architecture achieves high prediction accuracy with root mean squared error (RMSE) of (0,190 ± 0,024)%. Due to the use of a lightweight convolutional neural network (CNN) architecture, the model can be deployed on edge devices, such as onboard a planetary rovers. Moreover, the method is flexible, allowing for easy adaptation to other tasks—such as the analysis of more complex mixtures or inference based on multi- or hyperspectral imagery.

Keywords:

machine vision, machine learning, edge computing, in situ resource utilization, planetary rover

References

[1] S. Henley and R. Allington, “PERC, CRIRSCO, and UNFC: Minerals reporting standards and classifications,” European Geologist, vol. 36, p. 49, 2013.



[2] A. Sommariva, L. Gori, B. Chizzolini, and M. Pianorsi, “The economics of moon mining,” Acta Astronautica, vol. 170, pp. 712–718, 2020.



[3] A. Levison and R. Taylor, “Moon rocks and minerals,” Pergamon Press, p. 66, 1989.



[4] M. D. and W. R., “A geologic assessment of potential lunar ores,” NASA Ames Research Center, 1979.



[5] M. Hutson, “Notes of lunar ilmenite,” Lunar and Planetary Laboratory, vol. 465, pp. 1–2, 1989.



[6] H. U. Sverdrup and A. E. Sverdrup, “An assessment of the global supply, recycling, stocks in use and market price for titanium using the world7 model,” Sustainable Horizons, vol. 7, no. 1, p. 100067, 2023.



[7] https://www.scorpio.pwr.edu.pl/en/main. page/, “Scientific club of unconventional vehicles off-road.” (accessed on 24.04.2025).



[8] B. B. Carr, “Recovery of water or oxygen by reduction of lunar rock,” AIAA J., vol. 1(4), p. 921–924, 1963.



[9] D. Paige, M. Foote, B. Greenhagen, J. Schofield, S. Calcutt, A. Vasavada, D. Preston, F. Taylor, C. Allen, K. Snook, B. Jakosky, B. Murray, L. Soderblom, S. Loring, J. Bulharowski, N. Bowles, I. Thomas, M. Sullivan, and D. McCleese,

“The lunar reconnaissance orbiter diviner lunar radiometer experiment,” Space Sci Rev, vol. 150, pp. 125–160, 01 2009.



[10] C. Pieters, J. Boardman, B. Buratti, A. Chatterjee, R. Clark, T. Glavich, R. Green, J. Head, P. Isaacson, E. Malaret, T. Mccord, and J. Mustard, “The moon mineralogy mapper (m3) on chandrayaan-1,” Curr. Sci., vol. 96, 11 2008.



[11] Y. Surkov, Y. Shkuratov, V. Kaydash, V. Korokhin, and G. Videen, “Lunar ilmenite content as assessed by improved chandrayaan-1 m3 data,” Icarus, vol. 341, p. 113661, 05 2020.



[12] M. P. Charette, T. B. McCord, C. Pieters, and J. B. Adams, “Application of remote spectral reflectance measurements to lunar geology classification and determination of titanium content of lunar soils,” Solid Earth and Planets, vol. 79 (11), pp. 1605–1613, 04 1974.



[13] P. Lucey and B. Jolliff, “Lunar iron and titanium abundance algorithms based on final processing clementine uvvis images,” Journal of Geophysical Research, vol. 105, pp. 20297–20305, 08 2000.



[14] C. Pieters, J. Boardman, B. Buratti, A. Chatterjee, R. Clark, T. Glavich, R. Green, J. Head, P. Isaacson, E. Malaret, T. Mccord, and J. Mustard, “The moon mineralogy mapper (m3) on chandrayaan-1,” Curr. Sci., vol. 96, 11 2008.



[15] L. Palafox, C. Hamilton, S. Scheidt, and A. Alvarez, “Automated detection of geological landforms on mars using convolutional neural networks,” Computers and Geosciences, vol. 101, pp. 48–56, Apr. 2017.



[16] J. Li, L. Zhang, Z. Wu, Z. Ling, X. Cao, K. Guo, and F. Yan, “Autonomous martian rock image classification based on transfer deep learning methods,” Earth Science Informatics, vol. 13, pp. 951–963, 2020.



[17] L. Zhou, Z. Liu, and W. Wang, “Terrain classification algorithm for lunar rover using a deep ensemble network with high-resolution features and interdependencies between channels,” Wireless Communications and Mobile Computing, vol. 2020, no. 1, p. 8842227, 2020.



[18] F. Lv, N. Li, C. Liu, H. Gao, L. Ding, Z. Deng, and G. Liu, “Highly accurate visual method of mars terrain classification for rovers based on novel image features,” Entropy, vol. 24, no. 9, 2022.



[19] J. Razzell Hollis, K. R. Moore, S. Sharma, L. Beegle, J. P. Grotzinger, A. Allwood, W. Abbey, R. Bhartia, A. J. Brown, B. Clark, E. Cloutis, A. Corpolongo, J. Henneke, K. Hickman-Lewis, J. A. Hurowitz, M. W. Jones, Y. Liu, J. Martinez-Frıas, A. Murphy, D. A. Pedersen, S. Shkolyar, S. Siljestr¨om, A. Steele, M. Tice, A. Treiman, K. Uckert, S. VanBommel, and A. Yanchilina, “The power of paired proximity science observations: Co-located data from sherloc and pixl on mars,” Icarus, vol. 387, p. 115179, 2022.



[20] K. Edgett, M. Caplinger, J. Maki, M. Ravine, F. Ghaemi, S. McNair, K. Herkenhoff, B. Duston, R. Willson, R. Yingst, M. Kennedy, M. Minitti, A. Sengstacken, K. Supulver, L. Lipkaman, G. Krezoski, M. Henderson, T. Jones, B. Nixon, and R. Kirk, “Curiosity’s robotic arm-mounted mars hand lens imager (mahli): Characterization and calibration status,” tech. rep., Mars Science Laboratory, 06 2015.



[21] K. E. Herkenhoff, S. W. Squyres, R. Arvidson, D. S. Bass, J. F. Bell, P. Bertelsen, B. L. Ehlmann, W. Farrand, L. Gaddis, R. Greeley, J. Grotzinger, A. G. Hayes, S. F. Hviid, J. R. Johnson, B. Jolliff, K. M. Kinch, A. H. Knoll, M. B. Madsen, J. N. Maki, S. M. McLennan, H. Y. McSween, D. W. Ming, J. W. Rice, L. Richter, M. Sims, P. H. Smith, L. A. Soderblom, N. Spanovich, R. Sullivan, S. Thompson, T. Wdowiak, C. Weitz, and P. Whelley, “Evidence from opportunity’s microscopic imager for water on meridiani planum,” Science, vol. 306, no. 5702, pp. 1727–1730, 2004.



[22] T. Xu, B. Hapke, X. Zhang, Y. Wu, and X. Lu, “Micro-scale photometry of the moon using chang’e-3 panoramic camera (pcam),” Astronomy & Astrophysics, vol. 665, 07 2022.



[23] J. Zhong, J. Yan, M. Li, and J.-P. Barriot, “A deep learning-based local feature extraction method for improved image matching and surface reconstruction from yutu-2 pcam images on the moon,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 206, pp. 16–29, 2023.



[24] S. Els, N. Ageorges, M. Bogosavljevic, D. Kampf, Z. Ioannou, S. Amilineni, A. Salem Belal, A. Busoud, S. AlMulla, S. Mohan, P. Garg, B. Peng, C. W¨ohler, V. Lalucaa, C. Virmontois, S. AlMaeeni, and H. Almarzooqi, “The microscope camera cam-m on-board the rashid-1 lunar rover,” Space Science Reviews, vol. 220, 10 2024.



[25] V. R. Oberbeck, F. H¨orz, and W. Q. Robert Morrison, “Emplacement of the caley formation,” tech. rep., NASA, 1973.



[26] J. W. Head, L. Wilson, H. Hiesinger, C. van der Bogert, Y. Chen, J. L. Dickson, L. R. Gaddis, J. Haruyama, E. R. Jawin, L. M. Jozwiak, C. Li, J. Liu, T. Morota, D. H. Needham, L. R. Ostrach, C. M. Pieters, T. C. Prissel, Y. Qian, L. Qiao, M. R. Rutherford, D. R. Scott, J. L. Whitten, L. Xiao, F. Zhang, and O. Ziyuan, “Lunar mare basaltic volcanism: Volcanic features and emplacement processes,” Reviews in Mineralogy and Geochemistry, vol. 89, pp. 453–507, 12 2023.



[27] C. R. Steffens, L. R. V. Messias, P. L. J. Drews, and S. S. da Costa Botelho, “Can exposure, noise and compression affect image recognition? an assessment of the impacts on state-of-the-art convnets,” in 2019 Latin American Robotics Symposium (LARS), 2019 Brazilian Symposium on Robotics (SBR) and 2019 Workshop on Robotics in Education (WRE), pp. 61–66, IEEE, 2019.



[28] M. Seals, “On the robustness of object detection based deep learning models,” Master’s thesis, University of Tennessee, 2019.



[29] W. Dai and D. Berleant, “Benchmarking robustness of deep learning classifiers using two-factor perturbation,” in 2021 IEEE international conference on big Data (Big Data), pp. 5085–5094, IEEE, 2021.