Integrated Two Hopfield Neural Networks for Automatic LED Defect Inspection

整合雙霍菲爾類神經網路於自動化LED缺陷檢測

張傳育(Chuan-Yu Chang) ； 林思延(Chuan-Wang Chang) ； 張傳旺(Chuan-Wang Chang) ； 鄭慕德(Mu-Der Jeng)

《中國機械工程學刊》 29卷1期 (2008/02) Pp. 45-51

https://doi.org/10.29979/JCSME.200802.0006

摘要

晶圓缺陷檢測的目標在於偵測缺陷的晶粒。缺陷晶粒通常在電子顯微鏡的輔助下以視覺判斷的方式進行辨識。工程師需以肉眼檢查晶圓並手動標記缺陷區域，此舉往往無可避免地導入可觀的人事成本。本文提出一個包含有兩個霍菲爾類神經網路的自動化LED缺陷檢測方案來偵測晶圓影像中的缺陷晶粒。實驗結果顯示所提出的方法能有效的偵測發光二極體晶圓影像中的缺陷晶粒並具有良好的效能。

並列摘要

The aim of the wafer defect inspection is to detect defective dies and discard them. The defective dies were usually identified through visual judgment with the aid of a scanning electron microscope. Dozens of engineers visually check wafers and hand-mark the defective regions leading to a significant amount of personnel cost. In this paper, a complete solution which consists of two Hopfield neural networks is proposed to detect the defective dies of wafer image. The experimental results show the proposed method successfully identifies the defective dies on LED wafers images with good performances.

並列關鍵字

Hopfield Neural networks ； LED ； defect inspection