NASA Funds Collaborative Research to Further Mars Exploration Program
University of Houston computer scientist Ricardo Vilalta is joining forces with Dr. Tomasz Stepinski of the Lunar and Planetary Institute to develop new computational tools to characterize large portions of the Martian landscape.
Funded by a three-year NASA grant and led by Principle Investigator Stepinski, the project seeks to identify natural landscape structures, such as the inside of craters, the outside and inside rims of craters, the rims of inside craters, valley networks, and inter-crater plains.
Identifying these structures is important because rocks, minerals, and geologic landforms hold clues to past water activity on Mars, and understanding the history of water on Mars is a part of NASA's long-term Mars Exploration Program.
"Currently there is a lack of automated tools designed to assist planetary scientists with analyzing the surface of Mars, and only a small percentage of the data collected has been analyzed," says Assistant Professor Vilalta, a member of the UH Data Mining and Machine Learning Group.
"In fact, most current work is based on a method known as descriptive geomorphology, essentially consisting of narrating what is in a picture. The scientific community needs automated methods to look for complex patterns across Mars' surface."
Combining techniques from data mining, machine learning, and geomorphology, Vilalta and his research group are in charge of providing novel data analysis methods for the analysis of Mars' surface.
Stepinski is in charge of processing all data obtained from the Mars Orbiter Laser Altimeter instrument aboard the Mars Global Surveyor spacecraft. This data is subsequently used to construct global topographic maps of Mars in the form of digital elevation models.
"From a data mining point of view, the project is generating novel and computationally challenging techniques. For example, we are looking for new techniques to classify the surface of Mars with minimal expert intervention. Using a technique known as semi-supervised learning, we are exploiting information from very few regions of Mars and using that to label large portions of the planet's surface," says Vilalta.