The Characterization and Discrimination of Exoatmospheric Objects Through Multivariate Empirical Analysis
MASSACHUSETTS INST OF TECH LEXINGTON LEXINGTON
Pagination or Media Count:
The objectives of the effort were to investigate implications of data on 1000 radar signatures as to a the level of discrimination of signatures of the target class reentry vehicles from signatures of the non-target class other objects, and b to indicate the empirically derived decision rules and features of the signatures best for that purpose. The guiding philosophy was to develop an estimate of the best level of discrimination performance inherently possible, given the characteristics of the physical objects as evinced in the signatures, with minimal dependence on the physical model or on subjective inspection of the data. Methods used were the empirical analysis of multivariate data and, in particular, state-of-the-art pattern recognition techniques for feature creation, feature ranking, and discrimination algorithm development.