うねり曲線を抽出するローパスフィルタの研究(第3報) : 周波数領域法による高速高精度2次元ガウシアンフィルタの提案

概要

論文の詳細を見る
A 3-D roughness shape can be extracted by applying a 2-D Gaussian filter to 3-D measurement data to deduct the thus obtained 3-D waviness surface data from the measurement data. However, 8000×8000 measurement data, for example, requires a filter large enough to handle 1600×1600 data, meaning that several hours of processing time are needed. In addition, an end effect, namely ringing that occurs at a data end is also a big problem. As a way of solving these two problems, i.e. long processing time and the end effect, use of a spline filter is proposed. However, since it has a different transmission characteristic from that of a Gaussian filter, it will create a directional characteristic problem : filtering results differ depending on the calculation sequences in terms of x and y directions. If these four problems of processing time, end effect, transmission characteristic, and directional characteristic can be all solved, then a conventional Gaussian filter can be completely replaced with a new low-pass filter. This paper proposes to solve the above four problems the low-pass filter using the basic method of adopting a small-sized 12×12 filter, called a G-spline filter, to thinned-out data obtained by down-sampling of measurement data in the frequency domain.
社団法人精密工学会の論文
2005-09-05