Image Understanding Algorithm Development Environment for Hybrid Neural Network/Digital Computer Architectures

ORINCON proposes a fast-turnaround, high-payoff solution to the design and implementation of a Image Understanding Algorithm Development Environment (IUADE)f for Hybrid Neural Network/Digital Computer (HNNDC) architecture systems (e.g., per companion topic ARPA 94-080). Our Phase I effort will develop a detailed second-generation design for the IUADE based on an existing, proven HNNDC architecture. ORINCON's RIPPEN software combines a graphical user interface with a library of versatile software tools, and supports multiple traditional CPUs, neural network processors, and digital signal processors. RIPPEN funs on multiple platforms -- various computer workstations and ORINCON's PRIISM HNNDC/SPARC/i860 VME real-time architecture. ORINCON will make available an operational RIPPEN/PRIISM system, a sophisticated rapid-prototyping environment that fulfills all identified IUADE general requirements, early in Phase I as a baseline configuration and hardware/software testbed. We will coordinate our design and implementation activities with the Government and developers of companion IU application systems, conduct RIPPEN training, and analyze HNNDC IU requirements. We will use our IUADE testbed system in a rapid prototyping process that will incorporate our findings on requirements, architecture tradeoffs, and hardware and software technologies, resulting in an enhanced design by the end of Phase I. Anticipated Benefits: The IUADE will provide the ability to develop, implement, and test image understanding algorithms for end-to-end vision architecture systems. Applications include vision systems for autonomous vehicles, medical radiological and ultrasound image interpretation, intrusion detection/identity authentication for security purposes, quality control in manufacturing, and satellite image interpretation.