Vision-based automatic driving along innercity roads and across comple
x innercity intersections requires to detect and track road markings a
nd lane boundaries in order to determine the position and orientation
of the vehicle relative to the ground. The complexity of intersection
scenes and the disturbances in the detected contours enforce the use o
f model-based state estimation techniques. We recorded monocular image
sequences of complex innercity road intersections from a moving small
experimental truck in order to track intersection models using a Kalm
an filter. A scalar distance measure for the distance between image co
ntour points and model edge segments turned out to be advantageous for
the estimation process. This scalar measure is based on the simultane
ous exploitation of edge element location and direction,;thus using mo
re information about the image gray value variation than previously wh
en only the perpendicular distance between the edge element and the mo
del segment was taken into account. We report about the theoretical fo
undation of this approach, its implementation and experimental results
from several intersection sequences,;as well as detailed comparisons
with a previous, less sophisticated approach. These examples demonstra
te the kind of problems which are likely to occur more frequently at i
ntersection scenes than at those road scenes encountered while followi
ng a, more or less straight, uninterrupted highway lane with clearly m
arked boundaries. (C) 1998 Elsevier Science Ltd. All rights reserved.