身處在充滿人工智慧結晶的時代，我們視科技為理所當然，並享受著其帶來的便利與生活品質，然而在這項技術逐漸嶄露頭角之際，各類威脅也倚靠著科技滋長茁壯。政府2020年為復甦經濟所發放之「振興三倍券」於使用期限內曾傳出偽造事件，為了能精準判別整張有價證券真偽，本研究主旨為使用深度學習CNN (Convolutional Neural Network )，有效且快速辨別真偽振興三倍券微結構取樣影像組合，進而依此推測判別出整張紙券真偽，同時以減少訓練樣本數達到高辨識率為目標，取得最佳學習尺寸組合，最後歸類分析錯誤辨識微結構印刷類型並於原券定位，為此次研究目的。
首先將面額200元及500元之紙本振興三倍券掃描定義為掃描真券；與之複印後再次掃描為模擬偽券，後以尺寸32×32、64×64、96×96及128×128 pixels進行隨機局部不完全重複取樣，建立訓練及測試影像資料集，分組後個別輸入CNN模型訓練測試，得出辨識正確率與錯誤辨識影像於原券上之分佈。
實驗結果顯示，依照各組辨識正確率之比例及趨勢可成功推測判別整張振興三倍券真偽，且印證研究使用之CNN模型不需學習全尺寸之局部影像組合，僅訓練最大及最小尺寸之影像資料集，即可達到預期之顯著辨識成效；至於透過錯誤辨識分佈的統整，發現無論掃描真券或模擬偽券的局部取樣，所辨識的錯誤特徵皆有較高的比例集中於鈔券的凹版印刷處。
本研究提出一個不需藉由專業人士判斷有價證券影像，基於CNN模型即可有效辨別鈔券局部微結構真偽的方法，並以此實驗結果為基礎，未來可結合手機拍攝取樣，推測於拍攝指定距離範圍內之鈔券影像可精確判讀，達到更加便民與實用之效果。綜合上述，此研究不論是在產業界抑或是學術界皆具有一定程度之應用價值。

In 2020, “Triple Stimulus Vouchers” issued by government to revitalize the economy has been forged during the expiration date. In order to precisely identify the authenticity of entire securities, this paper proposes convolutional neural network (CNN) based deep learning algorithm to accomplish recognition of real and counterfeit Triple Stimulus Vouchers microstructure sampling image combinations, and according to the accuracy rate of recognition to speculate the authenticity of entire securities. Besides, this experiment aim to reduce training data to achieve high accuracy rate and get the best training size combination. By locating error recognition of microstructures in Triple Stimulus Vouchers can category and analysis different printing type of recognition. For the experiment, NT$200 and NT$500 scanned imagines of Triple Stimulus Vouchers are defined as real imagines, and simulated counterfeit imagines are produced by printing real imagines and scanned again to get. Train dataset and test dataset imagines are established by randomly partially overlap cropped with 32×32、64×64、96×96 and 128×128 pixels and then group them into 3 individual groups. Lastly, 3 groups directly feed to CNN model to obtain the accuracy of recognition and the location in the real and counterfeit Triple Stimulus Vouchers. According to experiment result, the authenticity of the entire Triple Stimulus Vouchers image can be successfully estimated and judged by the trend of the recognition accuracy of each group, and the largest and smallest size image data sets can be trained to achieve the expected significant recognition results. In addition, it was found that a high proportion of the error recognitions are concentrated in the intaglio printing of Triple Stimulus Vouchers. This research proposes a method based on the CNN model to effectively distinguish the authenticity and counterfeit of the local microstructure of Triple Stimulus Vouchers. Based on the experimental results, it can be combined with mobile phone shooting and sampling in the future. It is speculated that the securities imagines within a specified distance can be accurately interpreted. Based on the above, this research has a certain degree of application value whether in industry or academia.

Clement, J. L., & Risi, B. (1983). Laboratory examination of counterfeit and fraudulently altered documents. International Criminal Police Review, 368, 138-165.

Chambers, J., Yan, W., Garhwal, A., & Kankanhalli, M. (2015). Currency security and forensics: a survey. Multimedia Tools and Applications, 74(11), 4013-4043.

Centeno, A. B., Terrades, O. R., Canet, J. L., & Morales, C. C. (2019). Identity Document and banknote security forensics: a survey. arXiv preprint arXiv:1910.08993

Chowdhury, U. R., Jana, S., & Parekh, R. (2020). Automated system for Indian banknote recognition using image processing and deep learning. In 2020 International Conference on Computer Science, Engineering and Applications (ICCSEA) (pp. 1-5). IEEE.

Chauhan, S. P., Bahal, P. (2020). Introduction to Machine Learning. Emerging Trends in Big Data, IOT and Cyber Security, 115.

Dhankar, M., Walia, N. (2020). An Introduction to Artificial Intelligence. Emerging Trends in Big Data, IOT and Cyber Security, 105.

Fukushima, K., & Miyake, S. (1982). Neocognitron: A self-organizing neural network model for a mechanism of visual pattern recognition. In Competition and cooperation in neural nets (pp. 267-285). Springer, Berlin, Heidelberg.

Hinton, G. E., & Salakhutdinov, R. R. (2006). Reducing the dimensionality of data with neural networks. Science, 313 (5786), 504-507.

Hoang, V. D., & Vo, H. T. (2018). Hybrid discriminative models for banknote recognition and anti-counterfeit. In 2018 5th NAFOSTED Conference on Information and Computer Science (NICS) (pp. 394-399). IEEE.

Jadhav, M., kumar Sharma, Y., & Bhandari, G. M. (2019). Currency identification and forged banknote detection using deep learning. In 2019 International Conference on Innovative Trends and Advances in Engineering and Technology (ICITAET) (pp. 178-183). IEEE.

Kipphan, H. (Ed.). (2001). Handbook of print media: Technologies and production methods. Berlin: Springer.

Kamble, K., Bhansali, A., Satalgaonkar, P., & Alagundgi, S. (2019). Counterfeit Currency Detection using Deep Convolutional Neural Network. In 2019 IEEE Pune Section International Conference (PuneCon) (pp. 1-4). IEEE.

Lanceros-Méndez, S., & Costa, C. M. (Eds.). (2018). Printed batteries: materials, technologies and applications. John Wiley & Sons.

LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521 (7553), 436-444.

LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86 (11), 2278-2324.

Ludovic Arnold, Sébastien Rebecchi, Sylvain Chevallier, Hélène Paugam-Moisy. An Introduction to Deep Learning. European Symposium on Artificial Neural Networks (ESANN), Apr 2011, Bruges, Belgium. Proceedings of the European Symposium on Artificial Neural Networks (ESANN). <hal- 01352061>

Liu, Y., He, J., & Li, M. (2018). New and old banknote recognition based on convolutional neural network. In Proceedings of the 2018 International Conference on Mathematics and Statistics (pp. 92-97).

Lee, S. H., & Lee, H. Y. (2018). Detecting counterfeit bills and their forgery devices using CNN-based deep learning. In Proc. 13th Int. Multi-Conf. Comput. Global Inf. Technol. (pp. 16-20).

Modi, A. S. (2018). Review article on deep learning approaches. In 2018 Second International Conference on Intelligent Computing and Control Systems (ICICCS) (pp. 1635-1639). IEEE.

McCulloch, W. S., & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. The bulletin of mathematical biophysics, 5 (4), 115-133.

O'Shea, K., & Nash, R. (2015). An introduction to convolutional neural networks. arXiv preprint arXiv:1511.08458.

Prime, E. L., & Solomon, D. H. (2010). Australia’s plastic banknotes: fighting counterfeit currency. Angewandte Chemie International Edition, 49(22), 3726-3736.

Pham, T. D., Park, C., Nguyen, D. T., Batchuluun, G., & Park, K. R. (2020). Deep Learning-Based Fake-Banknote Detection for the Visually Impaired People Using Visible-Light Images Captured by Smartphone Cameras. IEEE Access, 8, 63144-63161.

Pilania, E., & Arora, B. (2016). Recognition of Fake Currency Based on Security Thread Feature of Currency. International Journal of Engineering and Computer Science, 5(7).

Rosenblatt, F. (1958). The perceptron: a probabilistic model for information storage and organization in the brain. Psychological review, 65 (6), 386.

Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323 (6088), 533-536.

Sufri, N. A. J., Rahmad, N. A., Ghazali, N. F., Shahar, N., & As’Ari, M. A. (2019, August). Vision based system for banknote recognition using different machine learning and deep learning approach. In 2019 IEEE 10th Control and System Graduate Research Colloquium (ICSGRC) (pp. 5-8). IEEE

Upadhyaya, A., Shokeen, V., & Srivastava, G. (2018, January). Counterfeit currency detection techniques. In 2018 8th International Conference on Cloud Computing, Data Science & Engineering (Confluence) (pp. 394-398). IEEE.

van Renesse, R. L. (1997). Paper based document security-a review. In European Conference on Security and Detection, 1997. ECOS 97. (pp. 75-80). IEE.

Wang, H., & Sun, L. (2020). Materials and Society: Functional Chemicals and Materials in Banknote and Security Paper. Accounts of Materials Research, 2(1), 1-6.

林明賢(2021)。摸一摸、看一看、轉一轉，認識振興三倍券紙本券與新臺幣現行流通鈔券防偽特徵。印刷科技，37(1)，1-34。

林政松(2016)。數位印刷技術與鑑定研析。印刷科技，32(4)，1-21。

周中興(2006)。2006安全印刷防偽材料考察。公務出國報告資訊網，1-26。

黃俊賢(2011)。鈔券最新印製技術趨勢研習。公務出國報告資訊網，1-39。

熊亮原、賴慕回、王禮豪、李昕蓉、陳思穎、潘珮宜(2015)。台灣印刷文物典藏網站與文創商品加值設計。中華印刷科技年報，209-226。

McKinsey Global Institute. Notes from the AI frontier: Applications and value of deep learning. Retrieved from https://www.mckinsey.com/featured-insights/artificial-intelligence/notes-from-the-ai-frontier-applications-and-value-of-deep-learning

TVBS新聞。快訊／首破獲偽造三倍券！ 檢警逮5人、主嫌聲押。Retrieved from https://tw.news.yahoo.com/%E5%BF%AB%E8%A8%8A-%E9%A6%96%E7%A0%B4%E7%8D%B2%E5%81%BD%E9%80%A0%E4%B8%89%E5%80%8D%E5%88%B8-%E6%AA%A2%E8%AD%A6%E9%80%AE5%E4%BA%BA-%E4%B8%BB%E5%AB%8C%E8%81%B2%E6%8A%BC-152533752.html

聯合新聞網。梅花浮水印難辨識！民眾陳情連超商店員都不敢收三倍券。Retrieved from https://udn.com/news/story/120974/4709594

2020 Findings from the Diary of Consumer Payment Choice. Figure 1: Share of Payment Instrument Usage by Year. Retrieved from https://www.frbsf.org/cash/files/2020-findings-from-the-diary-of-consumer-payment-choice-july2020.pdf