Representation and Recognition of 3-D Curved Objects
Journal
14 Annual Conference of Industrial Electronics Society
Pages
70 - 75
Date Issued
1988-10
Author(s)
Abstract
The objective of this paper is to describe some of the intermediate results on the 3-D curved object recognition and verification using 3-D range data. A representation of a 3-D curved object is extracted from a complete set of range data, as a global feature, a spine, of the object. The spine of an object is a unique features of a 3-D curve which can describe the shape of the object. The major axis of an object passing through the center of mass in the direction of the eigenvector with the largest eigenvalue is obtained by eigenvector fitting method using a set of range data for the boundary of an object. The object can now be considered as an accumulation of cross sections which are pe pendicular to the major axis obtained. The spine of an object is defined as a 3-D curved line connecting all the center points of these cross sections. As the spine of an object is unique, and invariant under translation/rotation it. is useful as a major feature to describe the 3 D curved object. Combined with other intrinsic features of a 3-D object such as volume, surface area, and elongation, the spine is used for measuring the deviation between a pair of objects. The method presented her has the advantages that the represr ntation is unique, and that it is less sensitive to the measurement noise as the description is based mainly on the global features. As these features are useful for measuring the similarity or deviations of objects, the features can be used for both recognition and verification purposes. © 1988 IEEE.
Subjects
A spine; Cross sections; Eigenvector fitting; Feature extraction; Invariance; Major axis; Pattern recognition; Range finder; Representation; Verification
Other Subjects
Agricultural robots; Computer vision; Feature extraction; Industrial electronics; Invariance; Object recognition; Pattern recognition; Range finders; Robotics; Verification; A spine; Cross sections; Eigenvector fitting; Major axis; Representation; Eigenvalues and eigenfunctions
Type
conference paper