Denoising Of White Light Interference Signal On The Metal Surface Using Bayesian Estimation

Since the characteristic information on the surface of engineering components is ofgreat value which provides a stationary solution for the deep research of formationprocesses and mechanical behaviors, the quantitative analysis referring to the fracturemicrostructure attracts more and more people’s attention. Generally, scanning electronmicroscopy (SEM) is used to get the microstructure on the object’s surface, but thetwo-dimensional images collected lack of longitudinal height information. In recentyears, according to the extremely complicated three-dimensional characteristics on thefracture surface, people turns to put emphasis on the three-dimensional way to exploreand obtain the microstructure on the fracture surface. Among the various methods usedto measure the microstructure, scanning white light interferometry has such technicaladvantages as simple structure, low cost, fast detection speed, nanoscale precision, etc.But in the process of many specific measurements, the white light interferencesignal collected contains a large number of noises as a result of the factors such asenvironmental condition, experiment equipment, which has a great influence on themeasuring accuracy of the detected signal. Therefore, the denoising pre-processing ofthe white light interference signal is particularly important, which can not onlyeliminate the noise in the interference signal, but also can highlight the characteristicsof the useful signal, and thus lay the foundation for the subsequent reconstruction ofthree-dimensional microstructure on the fracture surface.As for the complicated properties on the metal fracture surface, scanningwhite-light interferometry(SWIL) is introduced to acquire the white-lightinterferogram sequences, after which the white light interference signal in a singlepixel on the alloy fracture surface of30CrMnSiA is detected by dealing with thewhite-light interferograms with a MATLAB implementation. Moreover, owing to thestrong noise characteristics of the detecting signals, in this paper, we first apply a newdenoising method based on Bayesian estimation to process with the extremely weakwhite light interference signal on the sample fracture surface and furtherly simulationexperiment is done by comparing with the other denoising method (median filtering,mean filtering, fourier transform and wavelet transform). Finally, the signal denoisingalgorithm based on Bayesian estimation is employed to reconstruct the three-dimensional microstructure on the30CrMnSiA fracture surface which is thencompared with the original topography data by the methods of two-dimensional profileand power spectral density.Experimental results show that, according to the two denoising evaluationcriteria(signal denoising performance and signal spectrum), the denoising methodbased on Bayesian estimation not only improves the denoising performance of thedenoised white light interference signal, but also keeps the signal details and edgeinformation to the largest extent and contributes to effectively improve thedegeneration and distortion of white light interference signal, which turns out to be aneffective denoising method with easy implementation.