Abstract
最近幾年來，半導體工業一直是一個高度成長的產業。在半導體工業中所有製程都為自動化作業，而且大部份的檢驗作業也都是自動化作業。雖然目前已經有許多的自動化晶圓檢驗系統，但是大多都是設計做為製程監控而不是最終產品的檢驗。在這一篇論文中，我們發展了一個自動化晶圓檢驗系統。這一個系統可以檢驗三種主要的瑕疵：(1)比背景暗的瑕疵、(2)比背景亮的瑕疵以及 (3)焊墊的瑕疵。這一個系統所使用的影像處理技術包含了遮罩、樣版比對以及形態學的技術。在實驗的部份，我們使用了154個影像來測試這個檢驗系統的績效。實驗的結果顯示第一種瑕疵的檢驗率為95.45％、第二種瑕疵的檢驗率為98.7％而所有的11個第三種瑕疵都能偵測出來。此外、焊墊的偵測率為90％。Semiconductor industry is a fast growing industry in the recentyears.All the processing steps in the production line areautomatic,and most the inspections are automatic too.Althoughthere are many wafer inspection systems for process-monitoring,there are few automatic wafer inspection systems for finalchip inspection. In this paper,we develop an automatic waferinspection system.This system inspects three major defects,i.e.(1)the defect which is darker than background,(2)the defectwhich is lighter than background,and(3)the defect on pad.Theimage processing methods in this system are mask processing,template matching,and morphological filters.We use 154 imagesto test the performance of this system.The results ofexperiment show that the recognition rate of the first kind ofdefect is 95.45％,The recognition rate of the second kind ofdefect is 98.7％ and all the 11 defects on pad are found.Besides,the defection rate of pad is 90％.