A central axis and radius estimation method for torus object modeling

Abstract

Industries and homes are being swept up by the tide of automation which began with the development of computers and robots. Since the role of robots has become critical in automation industry, it has become an important issue for robots to behave and make judgments like humans when grabbing things. This study was conducted to accurately find the axis of an object when the object needs to be modeled for robots to make judgments and grabbing objects. Cylindrical objects are not difficult to estimate their axis if they have only one axis, but torus shapes have radii and axes in general. So we suggest a new method for a central axis and radius estimation for torus modeling based on: 1) 3D edge based segmentation (contour), 2) radius estimation using two surface's point normal vectors, 3) find the central axis using histogram analysis and 4) verification from projected 2D image and superquadrics equation. In this paper, we show torus modeling result from central axis and two radii estimation.