Abstract

DARPA’s Autonomous Land Vehicle (ALV) project is intended to develop and demonstrate vision systems capable of navigating in outdoor, natural environments. The major problems faced by this project, as we see it, are (1) developing a general-purpose vocabulary of models that are sufficient to describe most of the important landmarks that the vehicle will encounder, and (2) developing recognition techiques that will allow us to recognize these models from sensor data in both in a directed, top-down manner as well as in an unguided bottom-up fashion.

© 1987 Optical Society of America