Abstract

This paper describes a new, novel low computational-cost approach for recovering geometric structure and pose information of a 3D surface rotationally symmetric about an axis given an unorganized set of 3D range data which covers only a small patch of the surface. Self occluding contours are not seen here. All the parameters necessary to completely define the corresponding surface are estimated. The estimated parameters of the data patch consist of a line describing the axis and a set of parameters which describe the profile curve with respect to the axis. Global algebraic surfaces and local splines consisting of frustrums of cones are used as models for the surface estimation.