Abstract

With the advancement of technology and space exploration, research on the application of artificial intelligence in planetary data analysis becomes increasingly more significant. Autonomous control of space rovers allows for faster and more reliable exploration of Mars. This scientific paper is dedicated to the topic of terrain recognition on Mars using advanced techniques based on the convolutional neural networks (CNN). The work on the project was conducted based on the set of 18K images collected by the Curiosity, Opportunity and Spirit rovers. The training model benefits from the pretrained backbones trained for analysis of the RGB images. The project achieves an accuracy of 83.5% and extends the scope of classification to unknown objects when compared with related projects of Zooniverse and NASA's Jet Propulsion Laboratory scientific group.

Recommended Citation

Wicki, W., Burblis, W., Tkaczeń, M. & Demkowicz, J. (2024). Comparison of Deep Neural Network Learning Algorithms for Mars Terrain Image Segmentation. In B. Marcinkowski, A. Przybylek, A. Jarzębowicz, N. Iivari, E. Insfran, M. Lang, H. Linger, & C. Schneider (Eds.), Harnessing Opportunities: Reshaping ISD in the post-COVID-19 and Generative AI Era (ISD2024 Proceedings). Gdańsk, Poland: University of Gdańsk. ISBN: 978-83-972632-0-8. https://doi.org/10.62036/ISD.2024.27

Paper Type

Poster

DOI

10.62036/ISD.2024.27

Download

Share

COinS
 

Comparison of Deep Neural Network Learning Algorithms for Mars Terrain Image Segmentation

With the advancement of technology and space exploration, research on the application of artificial intelligence in planetary data analysis becomes increasingly more significant. Autonomous control of space rovers allows for faster and more reliable exploration of Mars. This scientific paper is dedicated to the topic of terrain recognition on Mars using advanced techniques based on the convolutional neural networks (CNN). The work on the project was conducted based on the set of 18K images collected by the Curiosity, Opportunity and Spirit rovers. The training model benefits from the pretrained backbones trained for analysis of the RGB images. The project achieves an accuracy of 83.5% and extends the scope of classification to unknown objects when compared with related projects of Zooniverse and NASA's Jet Propulsion Laboratory scientific group.