Model matching in robot vision by subgraph isomorphism

In this paper, we describe a systematic method for three-dimensional (3D) model matching in robot vision by using subgraph matching techniques. 3D objects are modeled as attributed graphs, called model graphs, where the nodes correspond to object vertices and the branches correspond to object edges. The 2D projections of a 3D object are modeled as subgraph isomorphisms of the object's model graph. Recognition is done by searching whether a 2D projection graph, constructed from the 2D projection of a 3D object, is a subgraph isomorphism of the object's model graph.