光学三维形貌测量中时频分析技术研究.pdf

摘 要 很多物理量可以被物体表面形貌的变化所反映。光学三维形貌测量技术具有 非接触性、高精度、高速性和自动化程度高等特点，在形貌测量领域得到了广泛 的应用。而傅里叶变换轮廓术作为一种主动光学三维形貌测量技术，因其只需要 一幅变形光栅图样来做全场分析的优点，而受到人们广泛关注。但是，传统的傅 里叶变换是一种全局频谱分析法，不适用于对复杂物面调制的非平稳变形光栅条 纹信号的分析，而且对噪声敏感。 本文不但回顾了傅里叶变换轮廓术中算法存在的缺点，而且研究了一类为了 改进傅里叶变换轮廓术中算法缺陷的非平稳信号时频分析算法，如窗口傅里叶变 换法、伸缩窗口傅里叶变换法和多尺度窗口傅里叶变换法。随后提出了窗口尺度 选取改进算法，该算法通过小波脊来提取信号的瞬时频率，然后控制逐点分析窗 口的宽度来保证窗口内信号的准平稳性。该算法在信号的频率分辨率和空间（时 间）分辨率之间达到一种更佳的调和，即使条纹信号被噪声污染时，仍然可以得 到很好的物面重建效果。 其次，研究了窗口尺度选取改进算法在非平稳信号时频表示当中的有效性， 并与现有的各种在时频表示中常用的时频分析算法相比较。比较结果表明，该算 法在运算复杂度与时频局域化能力综合考虑的前提下具有优势。 再次，研究了热门的经验模态分解方法在光学三维形貌测量当中的除噪应 用，光学三维形貌测量易受高频的背景噪声和 CCD 噪声的干扰。实验结果显示， 使用经验模态分解方法在去除噪声，提高信噪比方面具有优越性。 关键词：三维形貌测量；时频分析；小波脊；瞬时频率；窗口尺度选取改进算法； 经验模态分解 I Abstract Many physical quantity of an object can be reflected by changes of its shape. Optical 3-D shape measurement technology has been widely used in shape measurement because of its characteristics of non-contact, high accuracy, high speed and high automation. The Fourier transform profilometry as an active optical 3D shape measurement technology has been widely concerned because it need only one deformed grating image to make full-field analysis. However, conventional Fourier transform method is a global spectrum analytical method which is unsuitable to analyze non-stationary fringe signal that modulated by complex shape, and it is sensitive to noise. This paper not only review the shortcoming of the algorithm of Fourier transform profilometry, but also a class of time-frequency analysis algorithms for non-stationary signal has been researched. These methods are aim to improve the algorithm of Fourier transform profilometry, such as