微小被動電子元件之多面光學檢測

本論文係以微小被動電子元件，如電阻、電容，為對象，利用機器視覺（Machine Vision）和影像處理，對其外觀尺寸與表面瑕疵開發一套影像檢測流程，並配合自行設計的旋轉玻璃平台來研究自動光學檢測（Automated Optical Inspection，AOI）系統的實際運作。本論文提出的檢測流程包含了五大部分：動態取像、影像前處理（Image Pre-processing）、Hough Transform、形態學（Morphology），以及區塊分析（Blob Analysis）。本論文使用漸進式掃瞄（Progressive Scan）攝影機來擷取高速移動的元件，以避免因機台震動或是元件高速移動所造成的模糊影像，確保影像的清晰度與品質。由於攝影機擷取的是以等速通過鏡頭的元件，元件影像在畫面的位置與方向並非固定，必須加入前處理功能以定出元件位置後，才能進行瑕疵檢測。在影像前處理的部分，本論文使用對比度擴展（Contrast Stretching）與中值濾波器（Median Filter）增強影像及消除影像中的雜訊；接著對影像做二值化（Binarization）處理，並配合Hough Transform與區塊分析定位出元件的正確位置。利用同樣的方法找出元件電極端與本體的邊界，訂定出三個主要的檢測區域（陶瓷本體一區，兩電極端各一區）。在特定的區域內進行檢測，一方面用以減少檢測時的計算量；另一方面則用以避免因檢測區域過大而造成判斷上的錯誤。而為了避免光源因素（包含照射角度與強度等）造成元件的電極端產生非瑕疵的陰影，致使程式將其誤判為瑕疵，本論文利用形態學將電極端上的陰影或空洞處填補起來，盡可能使得兩電極端呈現出一個完整的區塊（Blob）。並使用區塊分析計算檢測區域內各個區塊的幾何特徵，如面積（Area）、邊界範圍（Bounding Box）等，來判斷元件是否合乎設定的標準。若其瑕疵在容許範圍內，則為良品；否則即判定為瑕疵品並進行瑕疵分類。本論文提出的微小被動元件檢測流程在取得元件各檢測項目的資訊時，即同時進行各項瑕疵檢測。若在檢測流程中找到瑕疵時，隨即判斷為不良品並做出分類的動作，有效地縮短每個元件的檢測時間。

關鍵字

被動元件；機器視覺；影像處理；自動光學檢測；瑕疵檢測與分類

並列摘要

This paper develops a procedure to inspect the basic dimensions and surface defects of miniature passive components, such as capacitors and resistors, by using machine vision and image processing. And we construct a system to simulate real-time operation of an automated optical inspection (AOI) system. The inspection procedure proposed here includes five principal parts: dynamic acquisition, image preprocessing, Hough Transform, morphological operation, and blob analysis. Firstly, we use a progressive scanning CCD camera to capture blur-free images from the rapid moving components under the lens. Because the orientation of a component and position of the component are not fixed, we have to locate the component image from the captured image before making any inspection by machine vision. Secondly, we perform image preprocessing to separate the component from background, and use Hough Transform to find angles of top and bottom edges of the component. Then we use the angles to correct the orientation of the component. Thirdly, we binarize the image to obtain blobs of component and electrodes respectively, which a blob (binary large object) is an area of touching pixels with the same logical state. Then we perform blob analysis to extract geometric characteristics of the blobs to inspect the dimensions of component and electrodes. In order to inspect the surface defects of body and electrodes, we have to find distinct defects, such as gap and hole, on them. Fourthly, we perform morphological operation on body and electrodes to fill tiny gaps and holes. Then we perform blob analysis to extract geometric characteristics of defect blobs and gauge sizes of those blobs to determine whether the component is good in surface defects inspection. Finally, we classify surface defects into two categories: electrode defects and body defects. Electrode defects include damaged and contaminants. Body defects include bright defects and dark defects. The inspection procedure this paper proposed proceeds dimensions and surface defects inspection while acquiring information for inspection. In other words, using this procedure to inspect passive component does reduce the inspection time for every component.

並列關鍵字

Passive Component ； Machine Vision ； Image Processing ； Automated Optical Inspection(AOI) ； Surface Defects Inspection and Classification

參考文獻

[1]. Douglas W. Raymond and Dominic F. Haigh, “Why Automate Optical Inspection?” International Test Conference, in Washington, DC, USA, Nov. 1997, pp. 1033

[2]. 曾俊洲，「被動電子元件製造業基本資料」，台灣經濟研究院產經資料庫，May 2004

[6]. Otsu N., “Threshold Selection Method from Gray-Level Histograms,” IEEE Transactions on Systems, Man, and Cybernetics, Vol. 9, No. 1, pp. 62-66, Jan. 1979

[8]. 徐偉邦，”運用影像分割技術於精密零件幾何形狀辨識”，元智大學，機械工程研究所，碩士論文，Jan. 2006

[9]. Serra J., Image Analysis and Mathematical Morphology, Academic Press, New York, 1982

國際替代計量

微小被動電子元件之多面光學檢測

全文下載

主題瀏覽