Representation and matching for a 2D vision system

Abstract

The thesis presents a high-level representation language for a 2D robot vision system that enables representation of objects with substructure, rich relationships between objects, and objects with regular, repetitive substructure. The language can represent both particular objects in the robot's field of view and classes of objects or collections of objects that can be used to classify particular objects. The thesis also describes the design and implementation of a matcher that can match descriptions in this representation language. The matcher could be incorporated into the robot's procedure interpreter, so that the robot would have a 2D "Eye of God" vision sensor.