Because of their speed and cost effectiveness, machines are continuing to replace "unskilled" human labor. If the United States is to maintain its competitiveness in the world markets, its machines must become still more sophisticated, replacing human labor in more complex or dangerous tasks. This transition would be greatly accelerated if machines could "see" like humans. The front end preprocessor for human vision is the retina. By parallel edge processing, it transfers to the brain, in real-time, the essential characteristics of the scene being viewed. The brain then interprets the image outline and initiates behaviors. This is a project for a very high-performance retina-like microelectronic processor called the Super Retina Analog Parallel Processor (SRAP). The SRAP is a high-speed analog parallel processor that outputs retina-like compressed-image information. It is faster than the human retina and in addition, could be used in the infrared as well as the visible spectrum. It is equivalent to hundreds of conventional digital processors working in parallel and is orders of magnitude smaller in size, weight, and cost. The research objectives are: 1) Define the imaging requirements for SRAP-relevant machine and robotic vision scenarios, isolating and emphasizing near-term commercial applications. 2) Translate these requirements into modifications of the baseline SRAP chip designs with as much general applicability as possible. 3) Verify the function and fabrication of the chip designs by SPICE simulations.
