應用類神經網路於球閘陣列基板之自動化表面瑕疵檢測

Automated Surface Defect Inspection of Ball Grid Array Substrates by Machine Vision System

林宏達(Hong-Dar Lin) ； 陳興倫(Hsing-Lun Chen)

《品質學報》 16卷1期 (2009/02) Pp. 43-60

摘要

爲因應電子產品朝向輕薄短小與多功能化的發展趨勢，其產品內部元件之設計亦朝向高精密度、小型化與高腳數化，而印刷電路板配合其主動與被動元件的發展，體積更顯著性縮小，其產品檢測準確度與精密度亦相對提升。而本研究針對印刷電路板中重要的球閘陣列(Ball Grid Array, BGA)基板，提出使用影像處理技術及類神經網路進行其表面瑕疵的檢測，藉由擷取待測BGA影像中物件的特徵值，與正常影像比對，再使用倒傳遞類神經網路瑕疵判別瑕疵是否存在。此方法不僅能克服因拍攝時所造成影像偏移或旋轉問題，而且解決影像經二值化後易造成物件縮脹的問題。實驗結果顯示，本研究所提方法的瑕疵檢測率高達99%，對一張1024×1024像素之BGA待測影像只需2.7秒的處理時間，極適合應用於線上即時之檢測。

關鍵字

表面瑕疵檢測 ； 印刷電路板 ； 球閘陣列基板 ； 類神經網路 ； 電腦視覺系統

並列摘要

Industrial inspection plays an important role in the manufacturing process of Ball Grid Array (BGA) industry. Since BGAs are basic components of many electronic devices, the quality of BGAs has a significant affect on the performance of many electronic products. Conventionally, visual inspection of BGAs is conducted by inspectors. The human inspections are subject to subjectivity, slowness, and inconsistency. To achieve high efficiency and effectiveness, it is essential to develop an automated BGA inspection system. Therefore, this research proposes an automated visual inspection system based on neural network approach for detecting surface defects of BGAs. This method applies the back-propagation neural network technique to feature comparisons of regions of interest (ROIs) with normal images to detect the surface defects in testing images of BGAs. Experimental results show the proposed method achieves defect detection rate of above 99%. The average processing time of a testing image of 1024×1024 pixels is close to 2.7 seconds. It is very suitable for on-line inspection applications.

並列關鍵字

surface defect inspection ； printed circuit boards ； ball grid array substrates ； neural network ； machine vision system

參考文獻

謝坤翰、蔡篤銘(2003)。球格陣列(BGA)基板表面瑕疵檢測。中國工業工程期刊。20(2)，125-138。

Batchelor, B. G.(1979).Streak and spot detection in digital pictures.Electronics Letters.15(12),352-353.

Bhuvanesh, A.,Ratnam, M. M.(2007).Automatic detection of stamping defects in leadframes using machine vision: overcoming translational and rotational misalignment.International Journal of Advanced Manufacturing Technology.32(11-12),1201-1210.

Chou, P. B.,Ravishankar, R.,Sturzenbecker, M. C.,Wu, F. Y.,Brecher, V. H.(1997).Automatic defect classification for semiconductor manufacturing.Machine Vision and Applications.9(4),201-214.

Edinbarough, I.,Balderas, R.,Bose, S.(2005).A vision and robot based on-line inspection monitoring system for electronic manufacturing.Computers in Industry.56(8-9),986-996.