The Research And Its Application On Blind Source Number Estimation On Depth-Resolved Wavenumber-Scanning Interferometry

The measurement of internal displacement or strain fields within a material has nowadays attracted worldwide attentions, widely used in quite a lot fields with great prospects, including in medical diagnostics and failure mechanisms. Multiple-wavelength techniques (MW) have been used as a depth-measuring tool in optical profilometry for a number of years. Wavelength scanning interferometry (WSI), which is one of the two basic forms of MW, is an extension of the traditional Phase-Shifting Interferometry(PSI). WSI records 2-D image sequences to describe 3-D profilometry of samples. Depth discrimination with multiple wavelengths has been used in optical profilometry for a number of years. The information for 3-D internal profilometry of samples is expressed by signals, therefore, the techniques above belong to Blind Source Separation (BSS). Like most BSS, Depth-Resolved Wavenumber-Scaning Interferometry(DRWSI) currently assume the signal number known, while the number is unknown in practical applications.This dissertation first introduces the background and the significance of DRWSI briefly, and then interpreted in terms of BSS.Research status of domestic and foreign at the beginning of the number estimation of BSS is then introduced. Next, some basic concepts of BSS the dissertation refers to are described, including BSS model and Principal Component Analysis (PCA). The DRWSI system the dissertation involved is also described.The following part is the focal point, Signal To Noise Equivalent Estimate (STNEE). The dissertation models for the system above in terms of BSS according to whose feature a method called STNEE is put forward. The basic idea of STNEE goes back to the PCA. In STNEE, the estimated signals and the estimates of noise main ingredients are comparedThe comparison between the estimated signals and the estimated main ingredients of noise are made, whose second derivative extreme the number corresponding to is regarded as the expected number. A detailed mathematical deduction is made and simulation figures are presented to show the execution of the algorithm. Furthermore, STNEE is both simulated and dealt with the data acquired by the DRWSI above to verify its performance. In simulation, it is to show how signal number and sampling point number to affect STNEE performance during different SNR. For experimental data, the algorithm show acceptable result. With the prior information that the signal number is discrete, a better result is gotten. STNEE, however, is based on second order statistics, while high order statistics information is not used. Meanwhile the algorithm require some prior information of sources per for better performance.This research of the thesis is closely linked with DRWSI and BSS, which has a certain instructional significance to the study of theory and algorithm of both.