Getting more from the scene for autonomous navigation: demo III XUV program

Abstract

The authors have applied a systems philosophy to the computer vision problem, and they have designed a system called O-NAV (Object NAVigation) that can harness all of the computer vision technology to date, and combine these approaches into one integrated system. In O-NAV, no one sub-component bears the burden of the problem. In other words, it is not expected that algorithms alone will solve the computer vision problem. If we choose effective sensing that inherently performs some level of scene discrimination at the phenomenology level, algorithms will be handed a partially analyzed scene before they ever encounter the raw image data. The algorithms have been designed to exploit an optimized processing hardware infrastructure, to maximize computation for the "real-time" application of autonomous robot navigation.