Robust Processing for Gaussian Regression Filtering of Engineering Surfaces

Gaussian filter (ISO 11562) is widely used to effectively separate the waveness and roughness of engineering surfaces. However, the inevitable involvement of uncorrelated form error, the surface singularity features and the existence of the boundary effect have influenced the practicability of Gaussian filter. To solve the above problem, a robust approach of modulated Gaussian regression filtering (RMGR) is proposed in this paper. It makes full use of the plasticity, non-integrated and low-pass properties of cubic B-spline. The robust estimation theory is introduced to conduct pre-processing on modulated Gaussian filtering with the help of calculus knowledge. Finally, the robust evaluation reference is obtained reliably in the whole measured area. The experimental results show that the presented approach can not only enhance the robustness of classical Gaussian filtering, but also perform the multi-scale separation of surface topography, which lays a good foundation for parameter and function evaluation for engineering surfaces including curved surfaces with a high degree of flexibility.