Feature-Based ATR with SAR/FLIR Sensor Fusion

We propose a two-pronged program for developing Air Force capability in SAR/FLIR sensor fusion for automatic target recognition (ATR). In the first component of this program, we will design, implement and demonstrate a sequence of increasingly sophisticated SAR/FLIR fusion systems, each tuned to maximize the joint benefit of these complementary sensing modes for a different level of algorithm complexity. The point of departure is a baseline SAR/FLIR fusion system that can be implemented with ATR techniques ready today, and which provides a performance level against which enhancements can be judged. The enhancements then considered involved SAR-assisted pose estimation, feature-level fusion and feedback strategies for FLIR-directed enhanced SAR processing. In the second, dual component of our proposed program, we will develop analysis tools to assess quantitatively the benefits of different levels of fusion and to predict theoretically the value of various potential sensor enhancements. The evolutionary nature of our overall proposed approach increases its short-term probability of success, and the steps we have already taken in this proposal to investigate a first-level enhanced fusion concept lead us to believe we will make a significant contribution to SAR/FLIR fusion technology in Phase I.