Sun Ho Lee, Tae Eun Kim, Jong Soo Choi

The KIPS Transactions:PartB , Vol. 8, No. 2, pp. 201-207, Apr. 2001
10.3745/KIPSTB.2001.8.2.201,   PDF Download:

Abstract

3-D object representation played an important role in the field of computer vision. In this paper, we propose part-based superquadric model as a coarse description of volumetric part of 3-D assembly part object. We used parameterized part-based superquadric model as a volumetric parameter model. This has the advantage that it has the capability to describe a wide variety of volumetric primitive shapes with a finite number of parameters including translation, rotation, and global deformation. Therefore, the representation for 3-D volumetric part object is possible through the superquadric shape recovery to extract superquadric parameters. The input for the shape recovery is three-dimensional range data points. Shape recovery is performed by a least-squares minimization (LSM) of the objective function for all range data points belonging to a single part of the object. After then we obtain the set of superquadric parameters representing each single part of a 3-D object. As a result, the overall superquadric model representation for a 3-D object is composed with these sets of each part's superquadric parameters. This approach is studied in experiments with synthetic and real 3-D data to prove robustness and usefulness in 3-D shape representation for assembly part object recognition.

Statistics
Cite this article