Image Processing Using Temporal Cellular Neural Networks

This proposal addresses the development of a compact, high-speed camera system that integrates a Charge-Coupled Device (CCD) solid-state imager and a Temporal Cellular Neural Network (TCNN) (using three-dimensional TC technology) to provide real-time image processing capability for scientific and engineering applications. The integrated CCD/TCNN camera provides high-speed image collection and processing, automatic Target Recognition (ATR), and target tracking. One camera implementation that will be investigated involves using the biologically realistic Hybrid Temporal Processing Element (HTPE), developed by and available exclusively to IRSI (patent pending), the HTPE is a VLSI-based spiking neuron model that has several characteristics which make it well suited for high speed image processing applications. The initial camera prototype will employ commercially available and custom VLSI components. The CCD imager, digital processor, and positioning system will be off-the-shelf and the remainder of the components will be custom VLSI. Also, in the initial prototype,m the VLSI components will be fabricated in a planar IC technology. This simplified design approach for the initial prototype will allow rapid proof-of-concept. Ultimately, the camera will be a fully-integrated system implemented in state-of-the-art three dimensional VLSI technology.