动态特征萃取之虹膜辨识系统.DOC

動態特徵萃取之虹膜辨識系統 An Iris Recognition System Based on Dynamic Features Extraction 研究生：李孟諭 Lee, Meng-Yu 指導教授：蔡玉娟 Tsay, Yuh-Jiuan 【摘要】 近年來，生物辨識系統(Biometric Recognition System)已經被廣泛的應用在個人身分識別上，因為利用人的生理特徵來做為身分識別的依據具有更高的安全性。由於虹膜具有高唯一性與高穩定性，而且這兩特性是作為身分識別的特徵所必須具備的，因此，以虹膜作為身分識別的特徵具有極高的應用價值。本論文採用中國科學院自動化研究所提供的CASIA虹膜資料庫，108隻不同眼睛，每隻眼睛有7張8位元灰階影像，共756張原始人眼影像，分兩個階段利用虹膜擷取裝置，在室內的環境下進行拍攝，研究過程中，依照實驗的需求針對CASIA虹膜資料庫進行分類處理，實驗的影像為44張虹膜完整且無雜訊之人眼影像。本論文的系統架構主要功能為影像擷取與前處理模組、特徵萃取模組、分類辨識模組。由於，環境影響造成虹膜影像擷取時的明亮程度不一，藉由動態之直方圖法強化其特徵輪廓，並透過實驗以獲得較佳的虹膜特徵萃取方法，提供倒傳遞類神經演算法進行分類辨識，本論文實驗之數據結果進行顯示採用的方法之可行性與辨識效果，可供後續研究作為參考。 HYPERLINK /search?ei=UTF-8p=%E8%AA%B2%E6%9C%AC 關鍵詞：虹膜辨識、生物辨識系統、身分辨識、識別、特徵萃取、倒傳遞類神經網路。 【Abstract】 In recent years, Biometric Recognition Systems have been extensively adopted in personal identity identification. Especially for iris, it possesses singularity and stability which are most important for personal identity identification. For this reason, it is much more useful to make use of iris as a discriminating feature. The research takes the CASIA Iris Database provided by The Institute of Automation of the Chinese There are three modules of this theist structure: Image-adopting and Pre-procedure Module, Identity Extraction Module, and Classify Identification Module. In this theist we are focus on: under iris image adopting process environment might affect the light, and we can enhance its outline feature though movement histogram equalization method, at the same time getting better iris adopting approach though experiment. We can also provide Back-propagation Neural Networks Algorithm to classify identification, analyzing and comparing our experiment results which can also prove that our approach is practical and recognizable, and will also provide further studies as reference materials. Keyword：Iris Recognition, Biometric Recognition System Back-propagation Neural Networks