The objective of this SBIR Phase II program is to test and validate a feature-based algorithm for Automatic Target Recognition (ATR) applications that was successfully demonstrated during the Phase I effort. This algorithm exploits Evolutionary Algorithm (EA) technology to fuse data provided by multiple battlefield sensors. In particular, the Phase II program shall process Ladar, IR, and TV sensor data recorded during the THOR data collection program. The sensor-fused system accepts feature vectors extracted from these different sensor images and outputs a target identification decision with a high probability of correct identification and a low false alarm rate. The proposed EA approach represents an innovative structure for both improving ATR performance and streamlining computations associated with the more computationally-intensive model-based techniques. The approach utilizes proven EA software that has been developed and successfully implemented for other ATR applications under several programs for the Army and Air Force. Army platforms that would particularly benefit from sensor fusion concepts are unmanned aerial vehicles that typically carry multisensor payloads (E0, IR, and/or SAR) for conducting reconnaissance, surveillance, and target acquisition (RSTA) missions.
Keywords: Data Fusion, Multisensor Fusion, Automatic Target Identitication, Features, Evolutionary Algorithm
