- 學位論文
影像處理技術應用於鈔券印刷瑕疵檢測之研究
The Application of Image Processing Techniques on the Printing Defects Detection of Banknotes
摘要
現今美、日、德等先進國家的鈔券印製產業中，已使用自動化方式進行檢測，而目前我國鈔券之檢測仍以人工目視為主。本研究目的將針對鈔券印刷上可能發生的瑕疵狀況，建立一套鈔券印刷瑕疵檢測系統，所檢測的瑕疵種類包括:油墨不均、尺寸長寬比、折角、套印走版、墨點、條痕、髒污、走墨及油墨脫落等。研究中利用影像處理技術，並經由鈔券定位、適應區塊互補、灰階模組建立及利用倒傳遞類神經網路分類出瑕疵種類，達到鈔券印刷良劣自動檢出。經實驗結果顯示本研究方法，其影像處理檢測時間約為0.395秒，比用人工檢測需花上4~8秒要快上許多，未來在應用上其技術尚可延伸其他有價證券的瑕疵檢測。
並列摘要
Today, the Automatic Optical Inspection (AOI) system is wildly used to inspect the quality of paper money in the developed countries like USA, Japan, and Germany. However, be using human vision for the check of banknote in our country. In this paper, focused on printing of banknote, an AOI system is developed based on the technology of machine vision. The types of defect can be printing-ink uneven, size aspect ratio, breakage, chromatography deviation, spots, strip, stain, peeled ink, ink overflowing. In this study, adaptation blocks complementary, gray modules, digital image processing techniques and uses back-propagation neural network to recognize kinds of the defects are using to develop an AOI system for the printing defects detection of banknotes. It is shown that the system only takes about 0.395 seconds for image process in standard of 4~8 seconds by human vision. So we proposed automatic inspection system is better than human vision. It is also expected that such developed system can be applied in flaw inspection of some portfolios.