Abstract
本篇論文提出一種指紋建檔策略希望能以良好的建檔檔案使得整個指紋系 統能有所改善。本篇論文亦以實驗手法來導出兩枚指紋間的相似性量度函 數。此函數將會被運用在建檔策略之中。 在建檔策略中，有三個要點必 須要注意：１．愈多的指紋樣本要有愈好的建檔檔案；２．指紋資料更新 策略；３．建檔檔案的品質。我們以配對指紋特徵點的平均位置方法作為 指紋資料的更新策略。而建檔檔案的品質可視為相似性量度後的分數值。 相似性量度函數將由實驗方式推導得之，從實驗中我們亦會得知在指紋特 徵點比對法中的有效特徵，進而幫助設計出更好的相似性量度函數。不同 的建檔方法在本篇論文將會作一比較。另一個參考的相似性函數也會與本 篇論文所推導的函數作一比較。我們將會發現當使用本篇論文所提出的建 檔策略與相似性函數時，系統會達到最好的效果。 A method of constructing the fingerprint database for improving the performance of identification system is proposed. A measurement function, which is a similarity measurement in minutiae matching between two fingerprints and used in the process of the database construcation, is experimentally deduced. The method of constructing the database, the enrolling process mentioned in this thesis, has the following focal points: 1. more patterns, better database; 2. update strategy; 3. database quality. The average location of feature point is the update way in this thesis. The database quality is corresponding to the matching score value. The measurement function is deduced from several experimental test. From the tests, the effective features in feature point matching will be understood, and it help to design a better discriminant function and improve the system performance further. Comparison between different enrolling method is done. A reference measurement function will be compared with the deduced function, too. The result is best when involving the new measurement function and the enrolling process proposed in this thesis.