Elsevier

Computer-Aided Design

Volume 33, Issue 1, January 2001, Pages 45-56
Computer-Aided Design

Analysis of difference fairing based on DFT-filter

https://doi.org/10.1016/S0010-4485(00)00062-2Get rights and content

Abstract

In this paper, even-order-difference fairing (EOD-fairing) of point-sequence curve (PS-curve) is analyzed using a discrete Fourier transform filter (DFT-filter). It is shown that EOD-fairing of a reflection-padded PS-curve is equivalent to DFT-filtering, and a DFT-filter corresponding to the EOD-fairing operation is derived. Based on the concept of filtering measure, a method of determining an “optimal” damping factor for EOD-fairing is proposed, and then the validity of the proposed method is demonstrated via experimental tests.

Introduction

A dense sequence of data points representing a smooth curve is often called a PS-curve (point-sequence curve). A PS-curve is usually obtained from an optical scanner or a coordinate measuring machine, but it may also be obtained as a result of surface–surface intersection or during NC cutter path generation [1]. PS-curves are subject to digitizing errors that need to be “filtered” by employing some fairing operations. One of the most straightforward PS-curve fairing methods is difference fairing.

A PS-curve representing a “y=f(x)type” curve is called an explicit PS-curve, which is given as a sequence of (xk,yk). A point-sequence obtained from a “line-type” laser scanner or coordinate measuring machine may better be treated as an explicit PS-curve (instead of 3D PS-curve) because only the “height-value” (yk) at each pre-planned “measuring-location” (xk) is measured. If Δ=xk+1xk for all k, we have a uniformly spaced explicit PS-curve.

The subject of PS-curve difference fairing has been studied by a number of researchers. Practical procedures for 2nd- and 4th-difference fairing are presented in Choi and Jerard [1]. For a 2nd-difference fairing of periodic PS-curves, Taubin [2] proposes a DFT-filter in which “scale-factors” are employed to control the shrinkage effects. It is also postulated that finite difference fairing operations can be analyzed by using a discrete low-pass filter [3]. For fairing binary sequences, Legault and Suen [4] introduces a “local weighted averaging method” that covers difference fairing as a special case.

In this paper, (1) difference-fairing of explicit PS-curves is analyzed by using a DFT filter, (2) a method of determining “optimal” damping factors for use in difference-fairing is proposed, and (3) fairing experiments are performed to validate the damping-factor determination method. More specifically, it is shown that even-order difference fairing (EOD-fairing) of a reflection-padded [5] PS-curve is equivalent to a DFT filtering, and then a DFT filter (it is termed an EOD-filter) corresponding to the fairing formula is constructed. Based on the premise that an optimal EOD-filter is the one that is as close to an ideal low-pass filter as possible, a method of determining an “optimal” damping factor is proposed. A set of 2nd-difference fairing experiments has been carried out.

The rest of the paper is organized as follows. Preliminary definitions related to difference fairing are provided in Section 2, and major properties of difference fairing together with a DFT filter (called the EOD-filter) are presented in Section 3. Results of damping factor analysis based on the EOD filter are presented in Section 4, and experimental results are analyzed in Section 5, followed by concluding remarks in Section 6.

Section snippets

Preliminary definitions

Introduced in this section are basic definitions related to (even-order) difference fairing of explicit PS-curves. Also given are basic definitions regarding curvature of PS-curves together with a list of major symbols used in the paper.

Definition 1 Even-order difference

For a uniformly spaced explicit PS-curve {(xk,yk)}k=0K, an even order difference (EOD) at yk is defined recursively as [1]:D2nyk=D2(n−1)yk+1−2D2(n−1)yk+D2(n−1)yk−1,withD0yk=yk.The EOD is called a 2n-th difference. It is valid for k∈[n,Kn] for a non-periodic

Properties of difference fairing

Presented in this section are major properties of EOD fairing. (1) An EOD-fairing formula is derived. (2) Basic characteristics of 2nd- and 4th-difference fairing are presented. (3) It is shown that non-cyclic EOD fairing of a reflection-padded PS-curve is equivalent to a cyclic EOD fairing of a mirror-extended PS-curve. (4) Cyclic EOD fairing is shown to be equivalent to DFT filtering.

Damping factor analysis

In the previous section, it has been established that EOD fairing of a reflection-padded PS-curve is equivalent to a DFT-filtering employing the EOD-filter (16). To be presented in this section is a procedure for determining “optimal” damping factor values for use in EOD fairing. The EOD-filter (16) becomes a 2nd-order difference filter if n=1, and a 4th-order difference filter if n=2.

First, in order to achieve a positive filtering, the magnitude of the EOD-filter Fm should be less than one:

Experimental analysis of 2nd-difference fairing

Presented in this section are results of fairing experiments using different values of α for PS-curves having “random noises”. Before presenting the experimental results, the noise pattern of an “actual” PS-curve is compared with that of an artificial PS-curve. Shown in Fig. 10a is the frequency distribution of the waves in the digitized PS-curve shown earlier in Fig. 4 (obtained from a commercial laser scanner). Fig. 10b shows the frequency distribution for a PS-curve having random noises. The

Conclusions and discussions

This paper provides a systematic analysis on even-order difference (EOD) fairing of explicit PS-curves. First, it is shown that EOD-fairing of a reflection-padded PS-curve is equivalent to DFT-filtering. More specifically, an EOD-fairing formula is derived and a DFT-filter (called an EOD-filter) for the EOD-fairing formula is obtained. Second, an over-filtering measure and an under-filtering measure are defined to represent the differences between the EOD-filter and an ideal low-pass filter,

Acknowledgements

The research was supported by KOSEF (Korea Science and Engineering Foundation) and MOST (The Ministry of Science and Technology).

Su K. Cho is currently a PhD candidate in the Department of Industrial Engineering at Korea Advanced Institute of Science and Technology. She received her BS and MS degrees in mathematics from KAIST in 1992 and 1994, respectively. Her research interests are in the area of surface modeling, surface inspection and reverse engineering.

References (12)

  • M.I.G. Bloor

    The smoothing properties of variational schemes

    CAGD

    (1995)
  • T.N.T. Goodman

    Automatic interpolation by fair, shape-preserving, G2 space curves

    CAD

    (1998)
  • B.K. Choi et al.

    Sculptured surface machining: theory and practice

    (1998)
  • Taubin G. A signal processing approach to fair surface design. Computer Graphs Proceedings, 1995. p....
  • R. Legault et al.

    Optimal local weighted averaging methods in contour smoothing

    IEEE Transactions on Pattern Analysis and Maching Intelligence

    (1997)
  • G. Smith

    Padding point extrapolation techniques for the Butterworth digital filter

    Journal of Biomechanics

    (1989)
There are more references available in the full text version of this article.

Cited by (4)

  • Tool-path generation from measured data

    2003, CAD Computer Aided Design
    Citation Excerpt :

    We assume that the measured data is free from digitizing errors. One can refer to some related results [18,19] to cope with digitizing errors. Fig. 1shows the overall approach of the paper.

  • Fairing the gamma: An engineering approach to sensitivity estimation

    2014, IIE Transactions (Institute of Industrial Engineers)
  • Adaptive data compression for massive scan data

    2004, Zhejiang Daxue Xuebao (Gongxue Ban)/Journal of Zhejiang University (Engineering Science)
  • Adaptive tool path generation from measured data

    2004, Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture

Su K. Cho is currently a PhD candidate in the Department of Industrial Engineering at Korea Advanced Institute of Science and Technology. She received her BS and MS degrees in mathematics from KAIST in 1992 and 1994, respectively. Her research interests are in the area of surface modeling, surface inspection and reverse engineering.

Byoung K. Choi is a professor of manufacturing systems engineering in the Department of Industrial Engineering at Korea Advanced Institute of Science and Technology since he joined KAIST in 1983. He received a BS from Seoul National University, a MS from KAIST, and a PhD from Purdue University, all in Industrial Engineering. His research interests are in the area of sculptured surface modeling, die-cavity machining, CAPP, system modeling and simulation, and virtual manufacturing.

View full text