Accession Number : ADA332739

Title :   Framework for Automatic Target Recognition Optimization

Descriptive Note : Final rept. 6 Nov 96-31 Oct 97

Corporate Author : CHARLES RIVER ANALYTICS INC CAMBRIDGE MA

Personal Author(s) : Ruda, Harald ; Snorrason, Magnus ; Shue, David

PDF Url : ADA332739

Report Date : 31 OCT 1997

Pagination or Media Count : 81

Abstract : We have designed a framework for the optimization of Automatic Target Recognition (ATR) algorithms. Successful ATR algorithms are complex, with non-linear components and feedback between components, and thus do not lend themselves to traditional analytical optimization methods. A prototype of the designed framework has been implemented with a visual programming interface that simultaneously aids design decisions and provides opportunities for improvements and optimizations. This framework is applicable to individual algonthms, groups of algorithms, and whole ATR suites. The framework can accommodate larger systems where the ATR algonthm is but one part; it is also possible to embed the framework into a larger system. We established concept feasibility in Phase I, which specified a design and implemented a prototype for the ATR optimization framework entirely in Java. The Phase I effort included a built-in ATR taxonomy to aid algorithm design and successfully demonstrated algorithm optimization.

Descriptors :   *OPTIMIZATION, *TARGET RECOGNITION, ALGORITHMS, AUTOMATIC, JAVA.

Subject Categories : Optical Detection and Detectors
      Active & Passive Radar Detection & Equipment

Distribution Statement : APPROVED FOR PUBLIC RELEASE