基於單張干涉圖之表面形貌與尺度量測術

Abstract

本論文針對微米與次微米級之表面尺度與表面形貌量測提出一套完整解決方案，並且具有以下貢獻：(1)對於處理不同類型的干涉條紋影像，提出分別基於分群原理與線掃描法之兩種條紋中心自動定位技術。(2)建立一套單張條紋影像的極紋萃取演算法，其中此演算法可用於分析密度不一的非對稱條紋影像。(3)提出一種具方向性的改良型相位還原方法，使得條紋影像所對應的相位分佈只需根據所求得之極紋以及相鄰像素間的相對相位差即可求出。因此，待測樣本之三維表面形貌僅需透過單張條紋影像便能被快速地重建。(4)當量測大尺寸球面樣本時，由於其條紋影像中只有中央區域的一些條紋是清晰的，周圍的條紋會因為分佈過於密集以致於模糊不清而難以辨識，習知的技術並不適用於此類情況。因此一種針對大尺寸球面光學元件之二維表面尺度量測技術被提出以解決這個問題，其中待測樣本之直徑可在次像素之精度下測得，同時其表面高度也僅需利用單張條紋影像與所測得的直徑值便可估算而得。相較於其他表面形貌量測技術，所提出之方法具有易於實現與處理速度快的優點，並且可以避免刮傷樣本表面。本文之方法亦經由應用於量測不同尺寸之光學元件作為驗證。

A total solution of micron and submicron level surface metrology and surface profilometry is proposed in this dissertation, which has the following contributions: (1) two automatic fringe centroid positioning techniques based on the concept of clustering and line scanning, respectively, are proposed for processing different types of fringe patterns. (2) An extremum extraction algorithm for a single fringe pattern is presented, and this algorithm can be applied to analyze asymmetric fringe patterns with varying density. (3) An improved directional phase retrieval method is proposed. The phase distribution of the observed fringe pattern is determined according to the obtained extreme fringes and the relative phase difference between adjacent pixels. Therefore, the three-dimensional surface profile of a test sample can be rapidly reconstructed based on only a single fringe pattern. (4) When measuring a large-sized spherical sample, only some fringes in the central area of the fringe pattern are clear, while the fringes in the surrounding area are difficult to be recognized due to too dense distribution, and the conventional methods are not applicable to this case. Therefore, a two-dimensional surface metrology especially for large-sized spherical optical components is proposed to solve this problem. The diameter of the test sample can be measured with sub-pixel accuracy, and the surface sag is further estimated according to a single fringe pattern of the test sample and the obtained diameter. In comparison to other surface profilometries, the proposed method has the advantages of easy implementation and low computation, and its noncontact property can avoid scratching the measured surface. The proposed method is also verified by measuring the optical components with different scales.