Feature Based Sensor Fusion Using Evolutionary Algorithms

This SBIR Phase I program is directed at developing and testing a feature-based system that exploits Evolutionary Algorithm (EA) technology to fuse data for several battlefield sensors. The sensor-fused system will accept feature vectors extracted from different sensor images and output a target identification decision with a high probability of correct identification. The EA approach represents an innovative structure for both improving Automatic Target Recognition (ATR) performance and streamlining computations associated with model-based (i.e., template-based) techniques. The approach utilizes proven EA software that has been developed and successfully implemented for other ATR applications for the Air Force and Army. The research will focus on electro-optic sensors, but will also encompass relatively long wavelength sensors, such as millimeter wave (MMW) that provide a relative degree of weather penetration but at the expense of resolution in the image, as well as synthetic aperture radar (SAR) sensors that provide good resolution and have excellent weather penetration.