期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：6
出版社：Journal of Theoretical and Applied
摘要：Fingerprint detection and recognition process is a critical task in many biometric systems. Hence, the accuracy of detection and matching fingerprint plays an important role in many research works. Many fingerprints sample are degraded through such other transformations. Therefore, there is a need to find a method that able to detect and match fingerprints sample over all other circumstances and give high accuracy rate. There are two approaches are commonly used, minutiae based and image based approach, minutia approach faced several problems when fingerprint samples are degraded or rotated which lead to poor performance result, image-based approach is seem that its very robust against such other degradation issues. In this paper, we proposed a novel modification to one of the most robust algorithms that used to extract the interest points, called SURF. These points also called feature are later used for matching process after eliminating outlier sets. The proposed system is tested on FVC2002 DB1, and FVC2004 DB2 according to set of performance evaluation metrics such as FAR, FRR, SMR, EER and accuracy, which shows our proposed system can successfully find matching between one fingerprint and other one (1:1) matching, as well as (1:M) matching when more than one fingerprint images may have related to the query image.
