Multi-Spectral Fusion Techniques

Data Fusion Corporation (DFC) proposed the development of radiometric invariants for automated target recognition (ATR) using thermal imagery in Phase I, together with SAR imagery in Phase II. This development will complement DFC's current efforts for exploiting geometric invariants for targeting and ATR. This effort exploits the concept of thermal invariants and thermal history constraints to recognize objects in infrared imagery. The work will also consider geometric invariants and other quasi-invariants for multi-sensor fusion and pose estimation. Particular attention will be devoted to creating a statistical theory of thermal invariants so their performance can be predicted for real-world situations. Judicious use of 3-D models will aid in confirming or discrediting particular hypotheses and in the off-line development of the invariants. Fusion of invariants-derived hypotheses of identification will rely on additive probabilistic accrual together with contingencies for establishing confidence intervals.