A phase i program is proposed for identifying, implementing, and benchmarking a set of algorithms in a parallel (hypercube) computing environment. Specifically, the program will be focused on classes of algorithms selected from computer vision and image processing as applied to avionics, specifically the problem of object representation, prediction, and matching. Algorithms in these areas have extensive data and computational requirements which make them ideal candidates for massively parallel computer implementations. The hypercube is quickly becoming a dominant architecture in the parallel computing industry due to its inherent reliability as well as its economy of design and production. In addition, the hypercube is software-reconfigurable into other supercomputing architectures such as rings, trees, and meshes. The choice of algorithms and architectures described in this proposal are very strong areas of expertise at perceptis corporation and the university of tennessee. The probability of success of phase i is high because of the extensive technical background of the investigators in the proposal, and also because all the hardware and software tools needed for this work are firmly in place
