以混合式檢測法為基礎作為印刷電路板自動檢測的實現

Abstract

印刷電路板在製造的過程中，為保證產品的良好工作性能，需要在每一個步驟加以檢測以達到產品的高良率性。傳統採用人工檢測的方式，因為電路愈做愈小、生產愈來愈快速的情形下，遭遇到不少困擾。解決這個 問題的方法在於求助於其他方式的檢測方法。於本文的開端，將詳細介紹人工檢測方法以外其它各種對於瑕疵 檢測的方法，尤其我們將著眼於影像視覺的檢測項目上，特別是檢測瑕疵項目的介紹。之後，對於現有各種影像檢測演算法也將做一個詳盡的探討。 有參考模型的檢測演算法往往存在耗費時間的問題，而無參考的檢測演算法在瑕疵檢測上卻僅能辨識有限的瑕疵情形。另一種稱之為混合式檢測法，它能夠含有上述兩個檢測方法的優點，本文的目的即在於提出一套能執行的混合式檢測演算法。於本文中，我們將詳細介紹電路影像取得後所採取的各種處理步驟。最後，以一些實驗來驗證所提的混合式檢測法之可行性並分析影響其檢測之相關因素為何。

PCB requires inspection in the process of its manufacturing, but human operators would encounter more and more troubles for nowaday manufacturing environment. Thus, PCB flaw detection equipments are needed. In the beginning of the thesis, we introduce all kind of flaw detection methods, especially focusing on visual inspection. Next,we make a survey about nowaday visual inspection algorithms. Algorithms of referential approaches usually have the problem of time consuming, while nonreferential approaches have the problem of their limitation on identifying only some kinds of defects. Hybrid methods have the goodness of both of them. In this thesis, we have proposed our hybrid inspection method. From the PCB image captured by a CCD, we clearly depict every image processing steps in the thesis. Finally, we test the method with some experiments to see if the method would work, and discuss the reasons that affect PCB visual inspection.