One of the commonest problems in artificial vision is how to identify
an object by comparing its characteristics with those of the models in
cluded in the recognition system database. Relevant characteristics in
clude properties both of the object's surface (size, shape, and reflec
tivity) and of its edges. In general, the solution to the problem depe
nds (1) on the type of illumination used and whether its distribution
across the scene is uniform or not and (2) on the surface characterist
ics (reflectivity, texture) of both the object and the background. In
this first phase of research we have tried out a particular signature
function: the distance of all points on the object's border from the o
bject's center