FEATURE EXTRACTION FOR VISION GUIDED ROAD VEHICLES

This paper reports the early results of an investigation conducted at the University of Bristol into autonomous road vehicle control. The emphasis of the work is on realtime vision, with the objective of steering a vehicle on well-defined roads at speeds up to 60 km/hr. This report gives an overview of the computer hardware and the lowlevel vision algorithms currently being explored. Our algorithms fall Mo two categories: bootstrap and real-time analysis. For the bootstrap mode we present a basic edgedetection and surface segmentation approach and an alternative method based on texture analysis. Our current emphasis is on real-time analysis techniques. Here we present only a single straightforward real-time algorithm, more complex approaches being given in companion papers.