Technical NoteValley Average of Lines (VAL) and of Directions (VAD) in 3D
Introduction
The concept of averaging is pervasive in Geometric Dimensioning and Tolerancing (GD&T) [1], [2] and is used to fit planes and lines to measured points.
In this note, we consider the problem of averaging a set of lines in 3D. The importance of our contributions may be illustrated by the following anticipated application: When defining GD&T for manufacturing quality control and for testing conformance for assembly [3], one may wish to compute, for a selection of machined parts, the “average” of the axes of a cylindrical feature estimated from acquired sensor data (such as a Coordinate Measuring Machine). The variability of these axes with respect to this average axis and the deviation of this average from the desired datum axis make it possible to distinguish between noise and systemic error. This distinction it not possible when only statistics on the deviation between the individual axes and the desired datum are collected.
Section snippets
Prior art
It is reasonable to expect an average of a set of lines to possess the following list of properties:
- •
Order invariance: The average should be invariant under permutations of the inputs.
- •
Parameterization invariance: The average should not depend on the parameterization or a particular choice of representation for the input lines.
- •
Independence from sampling: The average should not depend on a particular sampling of input lines.
It is not trivial to construct an average that satisfies these desired
Valley average of directions (VAD)
We first present a particular definition of the average of a set of directions (unit vectors) in 3D. We call it the Valley Average of Directions (VAD).
Other options for defining , such as a linear average or the geodesic squared-distance minimizer on the Gaussian sphere [8], may be suitable for some applications, but, none of these options lead to a formulation of an average of lines that satisfies the desired properties listed above. We show in Section 4 that VAD does.
Let be a
Valley average of lines (VAL)
We use the VAD to define our average of lines. Let be a set of lines in 3D, where line is defined by a point, , that it contains and by its direction, .
We define the average of as , where is the VAD of and where , with, being the sum of the squared distances from to the lines . In the above matrix–vector expression for , the quadratic terms have been encoded and expressed with the
Multi-line, order independent correspondence
Given lines in 3D, one may need a correspondence (common parameterization) between them that is independent of the order in which the lines are given. The VAL proposed here suggests two different solutions: CPontoVAL and CPfromVAL. For both solutions, we consider a parameterization of the average line, where is a point on the VAL and, for each , use it to define the corresponding point on .
Let return the point on line that is closest to point .
- 1.
CPontoVAL (Common
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (10)
- et al.
Review of dimensioning and tolerancing: representation and processing
Comput Aided Des
(1991) Modelling and representation of dimensions and tolerances: a survey
Comput Aided Des
(1992)A method for computer-aided specification of geometric tolerances
Comput Aided Des
(2013)- et al.
Average curve of n smooth planar curves
Comput Aided Des
(2016) - Levoy M, Hanrahan P. Light field rendering. In: Proceedings of the 23rd annual conference on computer graphics and...