鈔劵防偽瑕疵之光學檢測

Automated Optical Inspection for Security Features Defects of Banknote

黃識峰(Shi-Fong Huang)

指導教授 ： 吳明川

國立臺北科技大學/機電學院/自動化科技研究所/碩士(2007年)

https://doi.org/10.6841/NTUT.2007.00146

摘要

本研究針對印製鈔劵的安全防偽功能做瑕疵檢測，鈔劵在製造的過程中，有可能因為製紙、印刷或壓印安全線等過程發生瑕疵。本文乃是利用機器視覺的技術建立一套自動化瑕疵檢測系統。
然而，鈔劵的防偽功能辨識的方式也有所不同，使用的光源類型亦不同，本研究針對安全功能需要透射光源才能顯現出來的部份做瑕疵檢測，如圖案水印、數字水印、窗式安全線、盲人線及光影變化箔膜等，這些防偽功能在實際的印製過程可能會發生多種瑕疵，如未印製、位置上下左右偏移、歪斜、斷裂、凹陷等瑕疵。本文針對上述五項防偽功能，建立五個檢測模組，再利用影像處理的技術做瑕疵檢測，檢測的目標一方面確認各項安全防偽功能的正確外，另一方面檢測的過程必須迅速有效率。
研究中檢測鈔劵防偽功能是否有瑕疵所使用的影像處理的技術有投影法、二值化法以及相關係數法。本系統最後實驗測試結果檢出率達91.8%，整體平均檢測時間3.3~3.8秒。

關鍵字

鈔劵光學檢測 ； 防偽辨識

並列摘要

This study is aimed at the development of optical inspecting techniques for the detection of security features of banknote defects. During the manufacturing process of banknote, the defects may generate because of the material of paper, printing parameter, impression Window Threshold. This paper develops an automatic inspection system for making banknote process using machine vision.
There are different from recognizing methods and different light sources for security features of banknote. This study focused on security features of banknote that use back-lighted light sources, including Image Watermark, Numeral Watermark, Window Threshold, blind line, and Optical variable device stripe. During the manufacturing process of banknote, the security features defects are found. The defects can be no print, no inlay, slant, break, hollow … etc.. This paper designed five inspection models to cover five security features of banknote defects detection. The defects inspection system use digital image processing techniques. There are two objectives of defects inspection in this paper. The first objective is to confirm all security features of banknote. The second objective is to develop a fast and high efficiency defects inspection process.
We use image processing techniques for security features defects detection, including Projection Distribution, Threshold method, and Correlation Coefficient. The recognition correctness rate can achieve 91.8%. The processing time is about 3.3 to 3.8 seconds.

並列關鍵字

Banknote ； Optical Inspection ； Security Features Recognition

參考文獻

[6] 余忠潔，新的車牌定位方法，碩士論文，靜宜大學資訊管理學系，彰化，2002。

[8] 林建達，改良式相關係數法於瑕疵檢測之應用，碩士論文，元智大學工業工程與管理研究所，桃園，2002。

[9] 楊榮華，應用灰階共變異矩陣之多重指標於瑕疵檢測，碩士論文，元智大學工業工程與管理研究所，桃園，2002。

[12] J. R. Cowell, “Syntactic Pattern Recognizer for Vehicle Identification Numbers,” Image and Vision Computing, vol. 13, No. 1, pp. 13-19, February 1995.

[13] Chin-Teng Lin, Senior Member, IEEE, Chun-Lung Chang, Wen-Chang Cheng, “Application of One-Dimensional Machine Vision in the Textile Industry,” IEEE Transaction on Circuits and Systems, vol.51, No.5, pp.1024-1035, 2004.