Introduction

Automatic Target Recognition (ATR) systems detect and identify targets through pattern matching techniques, such as space-invariant transforms[1]/circular harmonic functions[2] or synthetic discriminant functions[3] (generalizations of matched spatial filters[4]). ATR enables such tasks as target acquisition for missile guidance systems, matching fingerprints or retina scans with their respective owners, and robotic delivery of items ordered in a restaurant to customers.

However, a common problem with ATR systems has to do with filters being created through manual selection of files and specific references therein. Not only is this method time consuming, but more importantly, often leads to poorly optimized filters with regard to the fact that more ideal references exist elsewhere for this targeting instance. The goal of this project is to develop a system capable of giving an ATR system access to a much more comprehensive selection of references, selected according to various parameters determined by the ATR system, yielding an increase in ability to locate and identify targets.

This implementation is centered around the requirements of target acquisition for the Image and Signal Processing Directive of Future Missile Technology Initiative (FMTI), allowing the ATR system to query the database to request target references meeting any conditions present in the database (such as distance from target, elevation, etc.)