傳統上，我們常用密碼來限制保全系統的存取權。然而當密碼外洩時，保全系統將會輕易的被破解。隨著科技的進步，生物特徵被用來改善這個系統保密的問題。在生物特徵辨識的領域中，指紋因為本身的唯一性和不變性而具有相當高的可靠度。此外，指紋辨識系統開發成本較低、使用方便等種種優越之處使得它已被廣泛的運用在資料保全系統上。 大部分的指紋辨識方法是以指紋的奇異點及紋線的端點、叉點的分佈位置、數目作為判別的依據。而擷取到的指紋影像與資料庫中的資料之間多少都會存在旋轉量與偏移量，因此會造成特徵點座標偏移的產生。為修正此偏移誤差，一般做法是將特徵點座標作多次固定角度的旋轉計算之後，再一一與資料庫中的特徵點資料做歐氏距離計算，此方法不僅無效率而且修正後的準確性也有進步的空間。 本論文提出一種新的比對方法來增進比對單元的效率，在一連串有效的影像前處理步驟之後進行特徵擷取，我們利用不同特性特徵點相對的關係作進一步的處理，並以經過特殊處理的特徵值作為兩枚指紋之間比對及判斷的依據，在演算法確立且軟體模擬成功之後，我們試著將提出的指紋影像辨識流程硬體化，利用Nios嵌入式處理器來實現個人身份辨識系統。 Traditionally, passwords have used to restrict access to secure system. However, security can be easily discredited in these systems when a password is leaked to an unauthorized user. With the growth of technology, biometric features are used to solve the problems in traditional identification system. In the field of biometric identification, fingerprints have one of the highest levels of reliability because of the uniqueness and invariability. Besides, fingerprint identification systems are also cost-effective, convenient and have been extensively used in data security systems. Most identification methods are dependent on the locations and numbers of ridge endings and bifurcations. The extracted fingerprint impressions usually have translation and rotation in different times which cause the coordinates of minutiae change. In order to align the position variance, the minutiae information is rotated a constant angle in several times and calculate the Euclidean Distance between query and database information. This method is inefficient and the accuracy needs to be improved. In this thesis, we propose a novel matching scheme to improve the efficiency of matching. After the pre-processing steps, the minutiae are extracted. We use the relation of minutiae as the information to identify fingerprint impressions. When the algorithm is simulated successfully, we implement it to build an automatic fingerprint identification system via a Nios embedded processor.
