The proposed effort focuses upon exploring the potential of a combination of neural, wavelet, and fuzzy logic processing to yield significant performance improvements needed for future seeker scene processors. Seeker scene processors must rapidly process images to detect and classify objects such as tanks on the ground, aircraft in flight, or ships at sea. The problem of detecting and classifying targets in thermal and visual images is complicated by factors such as sensor deficiencies, clutter, target size variations due to range, target aspect angle variations, sun angle and weather conditions. Moreover, targets may be partially obscured by smoke, dust, camouflage or shadows. A seeker scene processor must deal with real-time situations having largely unknown backgrounds and weather conditions. A simulation study will identify neural, wavelet, and fuzzy logic processing algorithms that can contribute significantly to the speed and accuracy of scene processing functions. Parallel versions of such algorithms will be investigated and a preliminary architectural design of a seeker scene processor capable of exploiting potential parallelism will be completed.
