Robust estimation in Gaussian filtering for engineering surface characterization

The reliability of reference datum is very important to characterize engineering surfaces. Gaussian filtering can effectively separate various surface components and yield the reference datum. However, when freak signals like scratches and pits are contained in the measured surface, the reference will be distorted. M-estimation is introduced to solve the problem in this paper. Several typical robust weight functions are adopted and compared with each other. Based on the comparison results, a novel ADRF robust weight function is proposed. In order to verify the feasibility of the new method, computer simulation based on the synthetic sinusoidal waveforms is used and a case study is conducted. The results show that the robust ADRF filtering not only possesses the efficiency of Gaussian filtering under normal conditions, but also enhances the robustness of Gaussian filtering under the conditions of abnormal interference.