NORMA eResearch @NCI Library

Enhancing Martian Surface Evaluations by Applying Multi-Task Machine Learning Algorithms to Satellite Images

Murphy, Daniel (2022) Enhancing Martian Surface Evaluations by Applying Multi-Task Machine Learning Algorithms to Satellite Images. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (2MB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (1MB) | Preview


Planning for manned missions to Mars in the near future is already well underway. However, the Martian surface topography is extremely complex and hazardous and requires accurate, detailed maps if these missions are to be successful. Various deep learning approaches are effective at mapping the surface for individual targets, but lack the ability to evaluate regions on multiple features. This research proposes a novel multi-task deep learning CNN to evaluate Martian regions based on two features: terrain classifications and crater detections. Three such multi-task model architectures, soft, firm and hard parameter sharing, are designed and compared to established single-task models. While the single-task model was found to outperform the multi-task for terrain classifications with recall and precision values of 41.95%, the multi-task model was found to have superior precision and F1 scores (5.52% and 2.15%, respectively) in crater detection. Hence, the novel approach allows regions to be evaluated on multiple parameters instead of single. Future work to improve the presented models will add more classification tasks to eventually be able to evaluate a given region across all relevant characteristics.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > QB Astronomy
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 23 Feb 2023 16:36
Last Modified: 02 Mar 2023 08:34

Actions (login required)

View Item View Item