Detecting Spots for NASA Space Programs Using Rough Sets

Abstract

This new method of detecting spots is based on the general concept of the rough sets. The lower approximation collects the set of objects which are assigned to a class in question without any doubt, and the upper approximation is composed of the lower approximation and the objects which are classified to the class with some uncertainty. In the initial step all objects detected are assigned to the class of spot-like objects (i.e. spot candidates). Subsequent refinements tend to extract spots with higher and higher degree of certainty, based on the lower approximation. Learning system is defined based on the rough sets making the learning phase automatic (by exploitation of the lower approximation) refining the set of candidates.

The work was under NASA contract No. NAS9-19100