Computer-Aided Geometric Design Based Three-Dimensional Object Representations For Computer Vision

Model-based recognition is one of the key paradigms in the field of computer vision and pattern recognition. However, at present there is an absence of a systematic approach to build vision models for a large class of three-dimensional (3-D) objects used in industrial environments. Computer-Aided Geometric Design (CAGD) permits the building of geometric models but does not support representations that are used in computer vision for object recognition.; In this thesis a CAGD-based approach is proposed to build representations and models that can be used in diverse applications involving 3-D object recognition and manipulation. There are two main steps in using this approach. First, the object's geometry is designed using a CAGD system, or its CAGD model is extracted from the existing database if it has already been modeled. Second, representations are developed from the above CAGD model and features are constructed possibly by combining multiple representations that are crucial in 3-D object recognition and manipulation. Examples of six CAGD-based representations are presented, which are: (1) surface points and normals, (2) polyhedra, (3) surface curvatures (4) extended Gaussian image, (5) generalized sweep, and (6) object decomposition and hierarchical representation.; The existing CAGD database is modified in novel ways and features are extracted in a systematic manner. Algorithms and experimental results are presented for building the interfaces between the CAGD models and their vision representations.