Recognizing Target Variants Using Transformational Adaptivity

Target variation presents a challenge for Automatic Target Recognition (ATR) based on fixed models or signatures. In Phase I General Intelligence refined and applied an extension of the map-seeking circuit (MSC) architecture to recognize articulations, plausible morphs and aggregations of target models. The extended MSC architecture implements a process termed “transformational adaptivity,” building on MSC’s capability in recognizing 3D targets from any viewpoint, tolerating clutter and distractors anywhere in the field of view including near the target, up to a substantial degree of occlusion. In Phase II GIC proposes to further extend the capability developed in Phase I to include potentially weapon-bearing targets of signficantly greater articulation and variation of surface form. While the mechanism refined in Phase I is tied to a particular kinematic model, the mechanism to be developed in Phase II, like human visual cognition, is extensible to far more complex configurations. In addition, a pre-processing stage will be developed to handle relative scene-camera motion and adaptively handle a much wider variety of input image conditions. A parallel objective of Phase II is to develop the image registration capability of MSC (which is the basis of handling relative scene-camera motion) for AFRL and GIC applications.