Discrimination

Even the most sophisticated countermeasures such as replica decoys and an antisimulated RV can be correctly discriminated based on their detailed physical properties; features that can be extracted using the radar and IR sensor assets of the BMDS using advanced discrimination algorithms. But to make the best use of these features, better methods of feature fusion are needed. In this proposal, Technology Service Corporation (TSC) presents a novel nonparametric Bayesian discrimination fusion technique that optimizes feature-based object discrimination while minimizing the potential reliance on assumptions about the independence of discrimination features. The new decision architecture is built on a novel classifier design that also has the potential to reduce the need for costly high-fidelity simulation during the classifier training process. The technique will be applied to the problem of discriminating the lethal re-entry vehicle (RV) from other types of objects in the ballistic missile threat complex using sensor features that estimate target shape, size and detailed motion state. The new classifier will be evaluated in the context of an attribute-based Decision Architecture (DA) of the type that is currently favored by the battle manager element (MDA/BC) to counter current and emerging threats including advanced countermeasures such as decoys and antisimulation.