Circle location from intensity and range data using the singular value decomposition

Abstract

Describes a circle location in three-dimensional space from a single view by fusing intensity and range data gained by stereo cameras as a part of our robotic disassembling project. An effective nonlinear "outlier" filtering and a linear 3D-circle estimation are briefly presented. The circle estimation is divided into two linear least-squares problems generalized for modeling higher dimensional data of what we call a "hypercircle". The singular value decomposition is intensively used for most computations throughout the paper.