激光波数扫描干涉检测的算法与实验研究

发布时间：2018-03-05 00:35

本文选题：光学干涉检测　切入点：相位测量　出处：《广东工业大学》2016年博士论文　论文类型：学位论文

【摘要】：激光波数扫描干涉技术通过透视测量复合材料内部的应变场,实现复合材料高精度无损透视检测,测量精度达到微应变级别,这是目前微米级缺陷无损检测的重要手段之一。然而受激光器波数输出范围有限影响,激光波数扫描干涉的深度z方向测量精度和轮廓分辨率均远逊色于传统的光学层析相干技术,并且复合材料内部介质的先验知识无法获取,导致不能有效分离出材料内部的干涉信号。另外,由于复合材料内部介质的反射光强过弱,造成内部干涉光强信号淹没于散斑噪声中。上述问题使得激光波数扫描干涉技术的实际检测效果远达不到理想要求。为解决上述问题,本论文提出在不增加激光波数扫描范围前提下,使用非线性最小二乘理论和空间谱估计理论提高激光波数扫描干涉的深度z方向测量精度和轮廓分辨率,并实现多表面干涉信号的盲分离,完成材料内部干涉信号的提取。同时,论文研究了基于非线性最小二乘理论的散斑噪声去除问题,提升了信噪比。论文主要研究内容和取得的成果如下：(1)使用复数域、实数域非线性最小二乘理论分别对干涉频谱和波数域干涉光强进行非线性拟合,分别建立了干涉频谱非线性最小二乘算法(Complex Number Least Squares Algorithm, CNLSA)和波数域干涉光强非线性最小二乘算法(Wavenumber-domain Least Squares Algorithm, WLSA)。CNLSA和WLSA可以去除傅里叶变换中的窗函数卷积模糊效应,取代傅里叶变换算法完成对激光波数扫描干涉数据的解调。实验结果表明在相同实验环境条件下,CNLSA和WLSA将测量系统深度轮廓分辨率由原来的δz分别提升到δz/1.8和δz/6。与此同时,相位测量精度因旁瓣压制能力增强而提高。(2)通过结合空间谱估计理论和相移干涉技术的数据解调原理,建立干涉信号相关矩阵谱分解及复数线性最小二乘理论(Eigenvalue Decomposition and Least Squares Algorithm, EDLSA),具体表现在：以空间谱估计理论为基础,通过深度分辨干涉信号自相关矩阵的谱分解,实现被测表面个数的盲估计；使用旋转不变技术求解干涉信号频率；在干涉频率已知情况下,将相移干涉技术中的实数域线性最小二乘算法发展到复数域,求解出干涉信号相位和干涉幅值,最终实现了激光波数扫描干涉信号的盲分离。实验结果表明EDLSA在激光波数扫描较窄的情况下,仍然可以根据CCD相机采集的光强值准确地盲分离出多表面干涉信号,且深度z方向测量精度和轮廓分辨率好于傅里叶变换算法。(3)在不破坏干涉显微x,y空间分辨率前提下,提出使用干涉频谱非线性最小二乘-傅里叶变换算法(Joint Fourier Transform and Complex Number Least Squares Algorithm, JFTLSA)对深度分辨散斑噪声进行降噪。具体表现为：充分利用深度z方向上的被测各个表面之间干涉信号幅值、频率以及相位耦合关系,改进干涉频谱非线性最小二乘方法,使非线性最小二乘待估参数维度显著降低,提高其在散斑噪声下的收敛性；针对空间卷绕相位建立噪声点定位方程,结合傅里叶变换算法,修正改进干涉频谱非线性最小二乘的卷绕相位发散点。深度分辨散斑干涉实验结果表明相比于傅里叶变换算法,JFTLSA可以使散斑噪声点数减少约20%左右。
[Abstract]:Laser scanning interferometry by wavenumber strain field measurement perspective composite materials, composite materials to achieve high precision nondestructive X-ray detection, the measurement accuracy of the micro level, this is one of the important means of the current micro defect nondestructive testing. However the laser output range of wavenumber limited effect, optical tomography laser scanning interferometry Z depth direction wavenumber the measurement accuracy and resolution are far inferior to the outline of the traditional coherent technology, and internal medium composite prior knowledge cannot be obtained, so it can not effectively isolate the interference signal inside the material. In addition, due to the internal reflection intensity of composite media is too weak, resulting in internal interference signals submerged in the speckle noise. Because of these problems laser scanning interferometry detection effect of actual wave number is not up to the ideal requirements. To solve the above problems, this paper proposed in Laser scanning range of wave number does not increase under the premise of using nonlinear least squares theory and spatial spectrum estimation to improve the laser scanning interference depth z wavenumber direction measurement precision and contour resolution theory, and realize blind separation of multiple surface interference signals, extract material internal interference signal. At the same time, the problem of removal of speckle noise in nonlinear least squares based on the theory, to enhance the signal-to-noise ratio. The main research contents and achievements are as follows: (1) the use of complex domain, real valued nonlinear least square theory of interference spectrum and frequency domain interference intensity were established by nonlinear fitting, interference spectrum of nonlinear least square algorithm (Complex Number Least Squares Algorithm, CNLSA) nonlinear intensity the least squares algorithm and wavenumber domain interference (Wavenumber-domain Least Squares Algorithm, WLSA.CNLSA and WLSA) You can remove the window function convolution Fu Liye transform fuzzy effect, replacing Fu Liye transform of laser scanning wave interferometric data demodulation. The experimental results show that under the same experimental conditions, CNLSA and WLSA measurement system of depth profile resolution from the original Z respectively to the delta delta z/1.8 and delta z/6. at the same time, improve the accuracy of phase measurement because the sidelobe suppression ability. (2) by combining the spatial spectrum estimation theory and the principle of phase shift interferometry data demodulation, establish interference signal correlation matrix spectral decomposition and plural linear least squares theory (Eigenvalue Decomposition and Least Squares Algorithm, EDLSA), specific performance in: spectrum estimation based on the theory of space, through the depth of resolution spectrum decomposition interference signal autocorrelation matrix, the measured surface number of blind estimation; rotational invariance techniques for solving interference channel use No interference in frequency; frequency is known, phase shift interferometry in real domain linear least squares algorithm to the complex domain, solving the interference signal phase and amplitude of the interference, finally realize the blind separation of laser scanning wave interference signal. Experimental results show that the EDLSA laser scanning in wavenumber narrow circumstances, can still be accurate to the blind separation of multiple surface interference signal acquisition CCD camera according to the intensity value and depth of Z direction measurement precision and contour resolution better than Fourier transform algorithm. (3) without damaging the interference microscope X, y spatial resolution under the premise of using the proposed nonlinear least squares spectrum Fourier transform algorithm (Joint Fourier Transform and Complex Number Least Squares Algorithm, JFTLSA) of depth resolution denoising speckle noise. The specific performance: make use of all the measured depth in the Z direction The surface between the interference signal amplitude, frequency and phase coupling relationship, improved spectrum nonlinear least square method, the nonlinear least squares estimated parameter dimension decreased significantly, improve the convergence of the scattered speckle noise; to establish a phase winding space noise equation, Fourier transform algorithm, a modified phase winding divergence point interference spectrum the nonlinear least squares. Depth resolution speckle interference test results show that compared to the Fourier transform algorithm, JFTLSA can make the speckle noise is reduced by about 20% points or so.

【学位授予单位】：广东工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN249

【相似文献】