Parallel Processing and Neural Networks for Real-time Target Classification

Period of Performance: 01/01/1988 - 12/31/1988

$509K

Phase 2 SBIR

Recipient Firm

LNK Corp., Inc.
6811 Kenilworth Avenue, Suite 306
Riverdale, MD 20737
Principal Investigator

Abstract

REAL-TIME CLASSIFICATION OF AIRBORNE TARGETS IS OF CRITICAL IMPORTANCE TO THE NAVY. WITH CONTINUING IMPROVEMENT IN SENSOR CHARACTERISTICS AND COMPUTER SOURCES, THIS GOAL BECOMES MORE REACHABLE. L.N.K. PROPOSES TO DEVELOP A PROTOTYPE AIRBORNE TARGET CLASSIFICATION SYSTEM USING RADAR CROSS SECTION DATA. THE SYSTEM WILL COMBINE AN L.N.K. MATCHING ALGORITHM WITH NEURAL NETWORK TECHNOLOGY TO PROVIDE FAST, ROBUST CLASSIFICATION. THE SYSTEM WILL ALSO INCORPORATE A STRUCTURAL PATTERN RECOGNITION APPROACH USING AN L.N.K. ARTIFICIAL INTELLIGENCE SEARCH TECHNIQUE. THIS TECHNIQUE ALLOWS FEEDBACK BETWEEN MODEL-DIRECTED AND DATA-DIRECTED SEARCHES TO OPTIMIZE THE EFFICIENCY OF THE CLASSIFICATION PROCESS. THE NEED TO DEVELOP A CLASSIFICATION SYSTEM WHICH CAN BE IMPLEMENTED IN HARDWARE WILL BE A PRIMARY CONSIDERATION IN PHASE II. ALL MAJOR COMPONENTS OF THE SYSTEM WILL MAKE EXTENSIVE USE OF PARALLELISM. BASED ON AN EARLIER AIRCRAFT CLASSIFICATION STUDY USING THE L.N.K. MATCHING ALGORITHM ON A RELATED FORM OF DATA, RELIABLE CLASSIFICATION APPEARS FEASIBLE. WE WILL TEST THE PROTOTYPE CLASSIFICATION ON SIMULATED RADAR RANGE PROFILES AND REAL DATA, AS AVAILABLE.