Using Partial Models for Localising Objects in Range Images.
Alan M. McIvor · 1998
: This paper describes a method for recognising and localising objects from a known class. Objects are modelled as a collection of simple geometric features and their spatial arrangement. Probability distributions on the parameters of the geometric features and their positions are used to handle the variability between members of the object class. Range images are processed to extract features (planes, spheres, cylinders) and these are matched against the model features using an interpretation tree based method to identify and locate the objects in a scene. Keywords: Range images, object recognition, partial models. 1