白光干涉三維輪廓重建-前段處理技術

White Light Interferometer For 3D Profile Reconstruction -Pre-processing Technique

李其聰(Chi-Tsung Lee)

指導教授 ： 陳傳生

元智大學/工程學院/機械工程學系/碩士(2005年)

摘要

由於工藝技術的提昇，元件加工對定位的精度、尺寸量測及製程工差的要求更為嚴格，因此光學檢測的技術便因應而生。而表面的量測技術可分為二維尺寸與三維輪廓的量測，其中以三維輪廓的量測最為複雜與困難。目前的量測方式不外乎是以雷射掃瞄、掃描式探針顯微鏡、顯微干涉輪廓儀為主，且量測精度及掃瞄面積皆有其一定限制，本論文將以白光干涉儀為研究平臺，直接掃瞄干涉條紋並開發研究更直接更快速的輪廓重建方法。
一般白光干涉大多藉由相位偏移技術(phase-shift technique)來求得影像的高度位置，透過手動調整物距使條紋出現在影像中，引入已知的相位到干涉場中，使圖形產生相位的改變。透過解相位的方式來取得待測物與參考面的相位差，從相位的解得，即可求出待測物體在載具上的高度值。但此技術需要很高的定位精度，且此方法受限於表面的高度資訊。若表面太過粗糙、高低起伏過大，當相鄰像數點的高度差大於 ，將會產生條紋在數量計算上及觀測上的問題，必須使用多波長記錄條紋數量，也需要更複雜的演算法來解釋影像，此外該技術也需要建立在條紋的搜尋上。
由於上述方法需考量定位精度，及條紋搜尋，因此本論文將藉由Depth From Focus對影像作對焦評估，利用對焦深度評估法來反算出待測物體於載具上的高度資訊。對焦深度評估法乃是透過搜尋法則及清晰度尺度運算，待找出對焦面及條紋後，在對焦面前後抓取一連串影像，引入Shape From Focus(SFF)及Focused Image Surface(FIS)將一整張影像分成數各小區域(window area)。藉由影像產生的清晰度值差異來找出各獨立區域對應的對焦高度值，而此值可視為該區域之平均高度，匯集全部的區域高度資料，即可完成3-D表面的輪廓重建。
藉由重建實驗，可以得知採FIS的作法會使相鄰區域的高度差值減少，細部結構流失，因此採區域計算的方式並不適合小尺寸、高度變化小的物件。在重建輪廓上，重現的輪廓細部結構的完整性取決於影像的擷取精度及取像設備的優劣，在動態取圖時，影像易受馬達及外在雜訊影響。因此基於影像處理的程式都必須加以考慮雜訊抑制，不然效果皆會未達預期。而深度評估的篩選條件也應更加理想，來適用各種圖形分佈的篩選。

關鍵字

相位偏移技術 ； 對焦深度評估法

並列摘要

Due to the state of art and the optical technology develop rapidly; the demands in the manufacture of the components are critical for the precision of the orientation, measurement of the dimensions and the tolerance in product, so the automatic optical inspection (AOI) technique is formed with the need. The classification of surface’s contour measurements can divide into 2D size and 3D contour, the 3D measurements are more complicated and difficult between two. Presently, the most popular equipment for surface’s measurement is laser scanner, scanning probe microscope, and micro optical interferometer. The differences between them are their resolution and the size of the scanning area. Our study is taking the interferometer as a research platform to reconstruct and recover the object’s contour by fringe of white light interferometer from scanning directly. Develop a higher and faster surface measurement technique.
Mostly, the white light interferometers use the phase-shift technique to get the exact height information. Adjusting the distance of the object lens make the fringes appear on the image. Using a known phase into the interfering field to makes the phase changes of the fringes. In this way, we can obtain the phase difference between the test object and the reference plane. Under the phase value getting, the height of the correspondent point to the reference plane can be calculated by this value. Although the Phase-Shift characteristic has great height precision, it is limited by the precision of moving orientation and the height of the surface. The height difference between two adjacent data points must be less thanλ/4 per, whereλis the wavelength of the light used. If the gradient of 3 D surface is greater thanλ/4 per detector pixel then the fringes will hard to be counted and observed. In this case, it needs to perform the measurement using two or more wavelengths. But the algorithm of image sharpness function will be complicated and time wasted. Besides this technique is also need to establish in fringes search.
According to the above-mentioned method，it has to estimate the precision of the orientation and the way of fringes search. Causative this factor, in this paper depth from focus is applied to auto-focus system in order to find the best focus point by using sharpness function and search algorithm. Once the focus plane and fringes are found, the Shape From Focus (SFF) and Focused Image Surface (FIS) will divide the full pixel image of object into several windows. Following is to find the optimal focus point of those individual windows by program sifting. And this value can take as the average height of individual window. To collect all the data points of window will reconstruct a full profile of the test object.
By the experiments of profile reconstruction, we know to use the FIS method will make the height difference decrease of the neighbor windows, loss detail structures. So this method is not suitable for small size, with height difference little of the object. In the profile reconstruction, the profile detail is decided by the resolution with grabbing images and the equipments. Under the dynamic grabbing, the image will easily be influenced by the motor motion and random noise. According to this reason, base on the image processing the program has to noise suppression, or the results will not ideal expected. Final, the program to sift the depth estimate still has to be improved to suit for more kinds curve in distribution.

Keywords： Automatic Optical Inspection (AOI), White light interferometer, Phase-shift technique, Focus depth estimate

並列關鍵字

Automatic Optical Inspection (AOI) ； White light interferometer ； Phase-shift technique ； Focus depth estimate