A system for associative object recognition of learned objects
by hierarchic template matching is presented. It is problem independent
to some extent. It is able to classify and recognize objects that are taken
by a TV camera in front of a known background scene. The hierarchic
steps are as follows: separation of object and background, object normalization,
classification, and individual recognition. For classification each
template is an archetype constructed from the learned objects of one class.
Three kinds of archetypes with respect to their classification efficiency are