This paper addresses automatic interpretation of images of outdoor sce
nes. The method allows instances of objects from a number of generic c
lasses to be identified: vegetation, buildings, vehicles, roads, etc.,
thereby enabling image databases to be queried on scene content. The
feature set is based, in part, on psychophysical principles and includ
es measures of colour, texture and shape. Using a large database of gr
ound-truth labelled images, a neural network is trained as a pattern c
lassifier. The method is demonstrated on a large test set to provide h
ighly accurate image interpretations, with over 90% of the image area
labelled correctly. (C) 1997 Pattern Recognition Society.