Advanced Optoelectronic Pulse-Coupled Pattern Recognition Systems

To develop faster and larger pulse-coupled neural networks (PCNN) for image processing, Bosonics will design and simulate a 64 by 64 neuron chip with optical inputs and outputs. PCNN have been shown to be effective at fast analog pre-processing of detector data, image transforms, and automatic target recognition. In Phase I Bosonics will 1) define the image processing application specifications and choose the appropriate optoelectronic integration technology; 2) design and simulate the neural cell functions, the optical system and candidate optoelectronic circuits; 3) design and simulate a 64 by 64 demonstration chip and design the test setups for delivery to MICOM. At the end of Phase I, the chip layout and simulated performance will prove the feasibility of the approach. In Phase II, Bosonics will fabricate and test the 64 by 64 optoelectronic neural array. The chips produced under this program will lower the cost of machine vision systms by providing an integrated detector/processing function as well as increasing performance by using computationally effective PCNN algorithms. Pulse-coupled neural networks have been shown be computationally efficient for low-level image processing. Thus, the chips developed under this program will have wide application in automatic target recognition, machine vision for automated manufacturing, and remote surveillance.