A self-supervised learning based approach to analyze Martian water–ice cloud properties for planetary atmospheric applications

Autor: John E. Moores, Gretchen Benedix, Shiv Meka, Daniel Marrable, Kevin Chai, Charissa L. Campbell, Christina L. Smith, Andrew L. Rohl
Rok vydání: 2021
Předmět:
Zdroj: Acta Astronautica. 181:1-13
ISSN: 0094-5765
DOI: 10.1016/j.actaastro.2020.12.041
Popis: Currently, Martian water–ice cloud properties, such as wind direction and angular wind velocity, are determined through manual analysis of atmospheric movies taken by the Mars Science Laboratory (MSL, Curiosity). These atmospheric movies, known as Zenith Movies (ZM), have a vertical elevational pointing which allows a direct measurement of wind properties from overhead clouds. However, acquiring this observation requires a significant amount of downlinked data volume which impedes on how often it can be taken. To combat this, an algorithm using Computer Vision (CV) and machine learning has been developed to calculate cloud parameters directly. To determine how well the algorithm performs, it has been tested on a previous data set from Campbell et al. (2020) that manually measured the wind direction and angular distance in ZMs. This data set had a variety of movies with different cloud features. When ZMs had strong features, the algorithm matched well with manual results which shows promising results. However, movies that had either lighting changes, multiple cloud decks or camera artifacts caused the algorithm to perform less well. Therefore the algorithm needs improving to more accurately measure these parameters over an assortment of conditions.
Databáze: OpenAIRE
Externí odkaz: