本篇論文結合了利用指紋上不同的資訊當作特徵來進行指紋辨識的幾種方法。這些不同的指紋資訊包含了指紋的特徵點、指紋紋線的資訊以及區域的指紋圖像。指紋的特徵點是指某條指紋上的終點或者是分叉點。這些特徵點所提供的位置以及角度便可以拿來進行指紋辨識。指紋紋線的資訊包含了在以某特徵點為中心的區域中,垂直於這些紋線結構的垂直線所經過的紋線數、這些經過的紋線到達所連接的特徵點之距離以及他們的曲率。而區域的指紋圖像是指我們會先將指紋以兩張不同的指紋所找到的特徵點為中心切成130x130的大小。之後我們利用這些切下來的區域,兩塊兩塊地來進行旋轉對齊後計算相似度。這三種不同的資訊會得到三個這兩張指紋的相似度分數,個別乘上權重值相加之後為最後的分數。在實驗中我們使用FVC2002資料庫進行實驗,實驗的數據會以equal error rate 簡稱EER為主,跟其他不同的方法做比較。所提出的方法由數據上來看有很好的結果並且可以勝過其他的方法。FVC2002的資料庫中,我們的方法平均的EER值為0.82,而傳統的特徵點指紋辨識的平均EER值則為8.12。
This thesis combines different fingerprint matching methods which are based on different features of the fingerprint image. The features on the fingerprint image that we used to matching are minutiae, ridge features and block skeleton image. The minutiae are defined as the ending points or bifurcation of the ridges, and the information used to match is their coordinates and orientation. The ridge features contain ridge count, ridge length, ridge curvature direction and ridge frequency. The ridge features are extracted in block around each minutia. In block skeleton image, we cut the skeleton image into a 130x130 block which takes a minutia as a center. The image registration finds the rotation angle of the cut blocks around the minutiae pair which is from the input fingerprint and query fingerprint. After rotation, we calculate the similarity of the two block skeleton image. The similarity scores of the three features summed with different weighting value as the final score of two fingerprints. Experiments are conducted for the FVC2002 database to compare the proposed method with other fingerprint methods on equal error rate (EER). The proposed method achieves better performance. The average EER value in the database of proposed method is 0.82 only and the average EER value of the conventional matching method using minutiae is 8.12.