Analysis and Characterization of Pattern Classifiers

"Frontier Technology, Inc. (FTI) and University of Florida (UF) propose to develop designs for automatically-generated statistical pattern recognition systems (GASPs) that can classify uncooperative targets among time-varying natural and manmadebackgrounds. We also propose to analyze the performance of the envisioned GASPs to:(a) covertly acquire feature data (e.g., statistical, spectral, and spatial cues) from target/background imagery,(b) apply multiple classifiers to target/background information to select probable target location and identity,(c) apply inferencing rules to disambiguate infeasible or contra-dictory classifier outputs.Pattern selection, key to successful system operation in mission- and threat-specific scenarios, will utilize Dempster-Schaefer the-ory and UF's powerful data fusion paradigm, Morphological Neural Nets (MNN).Phase-I will evaluate, extend and exploit FTI and UF's success-ful, DoD-sponsored R&D for dynamic pattern recognition and ATR to develop and test an efficient system design for target classifier output fusion and disambiguation. System design will includeanalysis of complexity and cost of potential hardware implementa-tions. In a Phase-II effort, we will use Phase-I results to drive candidate pattern downselection in FTI's DoD-supported TNE para-digm. MNNs and TNE have been proven highly successful in awide va-riety of recognition problems, thus we propose to analyze GASP sys-tem performance in realistic ATR scenarios. If successful, the proposed research will constitute a breakthrough in the solution of problems related to automatic target classifica-tion and pattern recognition. These problems occur extensively throughout both the military and commercial sectors: thepotential payoff is high. FTI will license or sell the solution to large aerospace companies for military applications. We plan to partner with a commercial company involved in law enforcement technologies, environmental monitoring and drug enforcementtechniques.C. KEYWORDSAdaptive Pattern Recognition, Neural nets (NN's), Morpholo