A kind of robust processing for Gaussian filtering mean line of surface profile

For the surface with grooves measurement, robust filtering is developing direction of traditional filtering, which is to try to minimize the distortion of mean line caused by the abnormal signal. This paper presents a new method for processing, which combines closely error theory with Gaussian filtering method. The abnormal signal is firstly judged« identified and corrected by 5σ criterion. Then Gaussian filter is implemented through moving average approximation method to obtain the mean line of surface profile. Finally, surface roughness is separated based on the original profile and mean line, and this method achieves robust filtering. It has simple algorithm, is easy to realize and can effectively restrain surface abnormal signal.