2024-03-29T04:38:43Zhttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_oaipmhoai:ipsj.ixsq.nii.ac.jp:000615322023-04-27T10:00:04Z01164:04619:05663:05664
自己参照に基づくパターン欠陥検査法Visual Inspection of Pattern Defect Based on Self-referencejpnhttp://id.nii.ac.jp/1001/00061532/Technical Reporthttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=61532&item_no=1&attribute_id=1&file_no=1Copyright (c) 2009 by the Information Processing Society of Japan和歌山大学システム工学部和歌山大学システム工学部日立製作所生産技術研究所日立製作所生産技術研究所淺海, 徹哉和田, 俊和酒井, 薫前田, 俊二本研究では,工業製品の検査画像から欠陥を検出する手法を提案する.画像の任意の位置のパターンについて類似したパターンを探し,その位置との比較を行い欠陥候補を特定する.さらに複数の同型な別の検査画像を用い,画像間の明るさむら・サブピクセル単位のずれを吸収しつつ高速に検出を行う.本手法により,繰り返しパターンとランダムパターンが混在する半導体基板画像から,小さく低輝度な欠陥を高速に検出できることを確認した.This paper presents a visual inspection method for detecting LSI wafer defects. Recently the density of LSI circuit patterns are rapidly increasing. Even in such a high density patterns, fast and accurate defect detection is still highly demanded. Some of the small defects can be analyzed only by using SEM, however, for faster detection without stopping the production lines, visual inspection using microscopic camera is necessary. The microscopic images of LSI circuit patterns can be affected by intensity variation caused by light interference in the thin layers. In most of the cases, the range of the intensity variation caused by interference is bigger than that caused by a defect. For detecting defects under such severe intensity variation, we propose self referencing image transform. Self referencing is actually a subtraction of local image from the most similar local image, which can be found by nearest neighbor search. By performing the self referencing image transforms on a defect-less die image and inspection die image, we can find the difference between these dies without affected by the intensity variation. We further extend this basic procedure so as to increase the accuracy. We tested the resulting algorithm on real images and confirmed its accuracy. AA11131797研究報告コンピュータビジョンとイメージメディア(CVIM)200929(2009-CVIM-166)2872922009-03-062009-08-18