Department of Computer Science at UH

University of Houston

Department of Computer Science

In Partial Fulfillment of the Requirements for the Degree of
Master of Science

Chaitanya Bagaria

Will defend his thesis

A FOUR-STAGE MACHINE LEARNING APPROACH
FOR AUTOMATIC IDENTIFICATION OF
CRATER COMPONENTS

Abstract

This work described in this thesis aims to further the goal of fully automated crater detection and analysis, by defining a framework for a system that automatically decomposes detected craters into their constituent parts. Such decomposition can be used for further calculation of other geometries of the crater. Using topography data and estimates of positions and sizes of craters, the system produces maps in which the craters' floor, walls, and rims are identified, and distinguished from all other terrain features in the input data. Earlier, a technique was developed that combined unsupervised pixel-based terrain classification with a segmentation mechanism to perform unsupervised terrain classification. The contribution from this thesis lies in modifying that technique to obtain terrain objects, which are then labeled using a supervised classification scheme. Finally, a post-processing filter is applied on the classification results to correct some mistakes and improve the visual appeal of the resultant map, according to the user's specifications. The experimental results are encouraging, with the system achieving more than 80% accuracy in identifying the components of test craters. Output maps are also produced for special categories of craters, such as multiple intersecting craters. The system is successful in identifying special regions such as those common to two or more craters, and producing maps as desired by the user.

Date: Monday, November 30, 2009
Time: 2:30 PM
Place: 550-PGH
Faculty, students, and the general public are invited.
Advisor: Dr. Ricardo Vilalta