Minimizing the Computation Time of Using a Robust Gaussian Regression Filter on Sanded Wood Surfaces

Abstract

Roughness of a processed surface has to be filtered to remove form errors and waviness. The mostcommon filter, the Gaussian filter, introduces distortions when used on some wood surfaces, whereas theRobust Gaussian Regression Filter (RGRF) does not. Unfortunately, the computation time when using theRGRF is increased significantly because the filter works iteratively and involves all profile data points in theevaluation. A modified algorithm that reduces the number of datapoints in the weighting window is proposedin this paper. The effect of the RGRF with a truncated window is compared to that of the RGRF withouttruncation on profiles of sanded wood and plastic and evaluated as an absolute error of Ra and Rtroughness parameters. Various weighting windows were tested and it was found that a window equivalent to1.25l (l being the filter cut-off length), gave negligible errors. The computation time is reduced significantlywhen the number of data points is limited by this weighting window.

Keywords

• Robust Gaussian Regression Filter
• weighting window
• sanded surfaces
• computation time