Abstract
在目前的半導體工業中有三種趨勢：(1)、基本元件愈來愈小。(2)、晶圓面積愈來愈大。(3)、晶圓層愈來愈多，製程愈來繁雜。這三種趨勢造成在檢驗晶圓時，人為檢驗已無法滿足一般瑕疵的檢驗，有必要利用一些機器來取代人為的檢驗，因此在半導體工業中，發展自動化的檢驗系統之重要性也愈來愈明顯。在本論文中，我們建立了一個晶圓表面瑕疵的檢測系統，這個系統能夠檢驗三個主要的瑕疵：規則區域的暗點、亮點及在鋁墊上的瑕疵，實驗結果表示在規則區域中暗點的辨識率是94.44％，亮點的辨識率是98.75％及在鋁墊上瑕疵的辨識率是94.44％，但在檢驗時間方面有點長，而在硬體上之改進以提升檢驗速度之構想在討論中提出。In the semiconductor industry,manual inspection is lessreliable due to the parts are becoming smaller and smaller.Thus, it is necessary use machine vision to replace the manualinspection.Automatic inspection has gained gradual emphasis inthe semiconductor industry. An automated inspection system ofwafer surface defects has developed in this study.This systemcan inspect three main defects,which are the darker defects,thelighter defects in the regular region and the defects on thepads.The experimental result shows that the recognition ratefor the darker defects in the regular regon is 96.25%.Therecognition rate for lighter defects in the regular region is98.75%.The recognition rate for the defects on the pads is94.44%.However,the inspection time is a bit too long.Suggestions about approachs to improve the speed of inspectionare proposed.