The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

Research Paper #728

Title:A Radial Basis Function Neural Network for Parts Identification of Three Dimensional Shapes
Authors:Borges,D; Orr,M; Fisher,RB
Date:Dec 1994
Presented:Accepted for presentation at the VII Brazilian Symposium of Computer Graphics and Image Processing, SIBGRAPI, Curitiba, Brazil, November, 1994
Abstract:The discrimination of volumetric pieces or parts of objects from range data is one key element for achieving 3-D object recognition. In this paper it is shown that previously segmented and acquired superquadrics from range data can be reliably mapped into a set of qualitative volumetric shapes (geons) by means of an RBF (Radial Basis Function) neural network classifier. We use a regularised RBF classifier and the results are shown to be both reliable and efficient in the context of range image understanding.

[Search These Pages] [DAI Home Page] [Comment]