Real-time Model Based Target Recognition System

Vexcel proposes to develop a model based automatic target recognition system capable of accurate, real-time recognition rates even in the presence of limited physical training data. Potential targets include any object for which accurate 3-D CAD models can be derived including, but not limited to: tanks, troop carriers, Humvees, aircraft carriers, frigates, etc. The proposed system is envisioned as a slimmed down, real-time version of the existing DARPA MSTAR ATR system, which is a feature based ATR approach that addresses the limited training data problem through the use of a Predict module which can produce expected SAR images given a target CAD model and a number of radar parameters. Features from expected images are then matched against features from the acquired image. Employing a proven ATR paradigm minimizes a substantial amount of risk, and also allows us to use the existing Predict module, as this would be impractical to develop during a Phase I feasibility study.