Researchers at SUNY (Stony Brook) request an eight processor, shared- memory system plus cameras, digitizers, and monitors for acquiring and displaying video images. This equipment will be used primarily in experiments to develop working vision systems. The main project is developing an automatic system to isolate faces from digitized visual scenes, to extract their key size- and orientation-invariant features, and to recognize them using a database of known faces. It emphasizes highly parallel algorithms, using low- level vision processing operators present in human visual systems. This work is producing efficient parallel algorithms for symmetry axis location and other critical vision operations. One result will be methods allowing real-time, automatic face recognition and tracking from video images using small arrays of processors running in parallel. Other projects will use this multiprocessor equipment: (1) to implement a real-time active-vision robot navigation system based on a binocular camera system controlled by a highly parallel algorithm, (2) to create a family of advanced environments with graphical user interfaces that support specification and design of concurrent systems, (3) to investigate the design and implementation of parallel algorithms for computer-aided design (CAD) tools for VLSI circuits and, (4) to experiment with methods of sharing memory and controlling global resources in large networks of computers.
