Low-Latency Embedded Vision Processor (LLEVS).

摘要

This report describes the efforts of a Small Business Technology Transfer (STTR) project to develop and analyze a Low Latency Enhanced Vision System (LLEVS). The LLEVS is an image processing engine capable of supporting image processing functions in helmet mounted and related mobile applications with less than 1 frame latency and at power levels less than 6 watts. The LLEVS must support a range of requirements the most difficult of which is comprised of two cameras and two displays of 5 megapixel resolution operating at 96 frames per second. The baseline LLEVS was predicated on using an imaging processing firmware suite that was developed to support helmet mounted systems. Empirical and estimated parametrics were derived from this suite and used to project the resources, performance and power projections for intermediate and ultimately objective levels for an LLEVS implementation. The data supporting the analyses has been acquired from physical measurements, development tools, simulations and estimates derived from the projection of measured parameters. The results have been used to define the path for an LLEVS capable of the objective performance. This report addresses the technology roadmap to assess the prevailing and projected technologies that will enable the design and implementation of the target processor capabilities.