| Abstract: | 近年來，在這多元化的社會中，人們生活方式也越來越複雜，在很多情況下，安全保密以及身分辨識是ㄧ項相當重要的課題。早期是使用密碼和鑰匙來當作個人的保密工具，印章和簽名來當作身分辨識，但是這類的工具存在著不方便性、易遺失、遺忘、盜用等問題因素。
由於人的身體特徵具有不可複製的特點，因此學者們都在探討，能否以生物特徵來當作個人身份識別的工具？在生物識別的領域中，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他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(核心點、三角點)及紋線特徵點(端點、叉點)的分布位置、數目作為判別的依據。而若能將指紋流向場精確的求得，那要擷取出特徵點以及核心點將是ㄧ件容易的事情。因此許多學者都提出了計算流向的方法，最常用的方法不外乎Stock與Swonger所提出的Slite-Sum，以及利用梯度方向來得到的指紋流向。
一般的指紋影像處理方式都是先將指紋影像經過背景去除、正規化、濾除雜訊、取指紋流向、流向的修正、Gabor濾波二値化、特徵點的搜尋與擷取以及最後的建檔與比對。但遇上龐大的指紋資料庫時，整體的比對效率是最常為人所詬病的，故本文提出ㄧ種新的指紋流向演算法(模糊聚類)來計算指紋流向場，以便更精確的擷取指紋中心核心點及修正指紋的偏移量與旋轉量，最後利用計算流向場所得到的歸屬度去做建檔或利用歐式距離的計算來做指紋資料庫比對，這將會大幅提升指紋辨識精確度以及未來實現在指紋辨識硬體方面的效率與準確性。
Now, people’s life are getting more and more complicated in this diversified society. In many situations, keeping secret safely and identifying recognition are quite important. In early days, people kept secret by using secret code and keys and recognized identifications by seals and signatures. But these ways are inconvenient and easily lost, forgotten and stolen.
Because the characters of human bodies can not be copied, scientists have explored the possibility of biological characteristics as a personal identification tool. In the field of Biometric recognitions, fingerprints because of their own uniqueness and invariance which have very high reliability. And it has widely used in protection systems because of the advantage of the lower development cost.
Most identification methods depend on the locations and numbers of ridge endings and bifurcations. If we can get the fingerprint flow field exactly, it will be easy to find out the minutiae and singular points. As the result, there are many scholars propose many ways to calculate flow. The ways mostly used are Slite-Sum methods proposed by Stock and Swonger. The fingerprint flow is got by using gradient.
Most of the fingerprint verification systems would including normalization, noise filter, calculate fingerprint flow direction, Gabor filter, detect the positions of the singular points and minutiae, as a basis for the later verification. But, if we have a huge database of fingerprint, the whole efficiency of compare is the most frequently denounced, therefore, here is a new fingerprint flow algorithm bring up in this thesis to compute the fingerprint flow field, for the purpose of selecting fingerprint core more accurate and correcting the offset and rotation of the fingerprint, in the end, use the membership grade by calculate flow field to set up file or use the compute of Euclidean distance to build up the compare of the fingerprint database, that will promote the accuracy by a wide margin and accomplish the efficiency and the accuracy in the hardware of the fingerprint identification in the future. |
