請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/59313
標題: | 基於小波轉換與機器學習的簽名真偽辨識系統 Handwritten Signature Verification System Based on Wavelet Transform and Machine Learning |
作者: | Ming-Ying Tsai 蔡銘穎 |
指導教授: | 曹恆偉 |
關鍵字: | 簽名辨識,手寫辨識,特徵擷取,電腦視覺,機器學習, signature recognition,handwritten recognition,feature extraction,computer vision,machine learning, |
出版年 : | 2017 |
學位: | 碩士 |
摘要: | 過去數十年來,手寫簽名辨識被廣泛且深入的研究,其應用包括金融業、
信用卡驗證、保全系統等等。一般來說,手寫辨識系統被分為兩個部分—線上 辨識以及離線辨識,線上辨識需要觸控筆及平板電腦等工具擷取動態手寫資訊, 而離線辨識則透過掃描器將手寫資訊轉換成圖片檔等靜態格式。 由於在本國,司法機構在針對偽造簽名辨識的問題上,並未有一套科學系統上的驗證,仍是以人眼透過視覺觀察簽名的各種資訊,諸如筆順、力道、弧度等來判斷是否為偽造簽名。本論文透過近年來快速進步的電腦視覺以及機器學習技術,提出一套系統化,並且可靠的方式,可在少量的資料量下,成功地將簽名分辨出是否為偽造簽名。 本論文首章為論文簡介,而第二章以及第三章為背景知識介紹,從第四章開始為本篇論文的主要貢獻,也就是系統架構設計,包含簽名影像的前處理、轉換、分類、測試等,並在第五章提供不同參數、不同模型的模擬比較,末章則為結語與未來展望。 In the past couple of decades, techniques for handwritten signature verification have been thoroughly studied and put into practice in various systems, which include security systems and financial field where credit cards verification is much needed. Generally speaking, handwritten signature verification system can be categorized into two kinds, online and offline verification. The former requires stylus pens and tablet computers to capture dynamic signature information, whilst for offline verification, a scanner is used to turn handwritten information into static formats such as image files. Due to the fact that there is not yet a scientific system for signatures verification upon forgery problems in the Taiwan judicial system, the procedure is mainly carried out through trained human eyes, with which validate the signatures with different unique cursive writing styles. In this thesis, a more systematic and dependable method on signatures verification is proposed. With the rapid development of computer vision and machine learning, it is possible to identify whether the signature is forged even under limited data. The layout of this thesis is as follows: The first three chapters give a brief introduction and provide background knowledge in the field of handwritten signature verification. The fourth chapter focuses on system architecture design, including data preprocessing, transformation, classification, and testing. Chapter five present the comparison with different parameters and simulation models. And finally, the last chapter wraps up the thesis with discussion and future work. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/59313 |
DOI: | 10.6342/NTU201701300 |
全文授權: | 有償授權 |
顯示於系所單位: | 電信工程學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-106-1.pdf 目前未授權公開取用 | 2.38 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。