Two-stage Fingerprint Recognition System Based on Robust Binary Invariant Feature and Multiple Image Quality Metrics

• Main Authors: Shao-En Lee, 李紹恩
• Other Authors: Jing-Ming Guo
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/79247064832683425609

碩士 === 國立臺灣科技大學 === 電機工程系 === 105 === With latest technological advents, biometrics recognition is becoming ubiquitous to identify individuals based on pshyiological and behavioral charactersitics. Specifically, in handheld devices, the fingerprint recognition is prevalently adopted and thus identification rate plays a critical role for commercial deployments. This paper proposes a novel fingerprint recognition system, which is involved with a two-stage identification strategy. To begin with, the paper proposes a Robust Binary Invariant Feature (RBIF) utilising adaptive threshold strategy which extracts features from the accelerated segment test. To obtain the rotation-invariant feature points, the gradient orientation is calculated and utilized to unify angles of features. Subsequently, Multiple Image Quality Metrics (MIQM) are used to ensure the veracity after the feature matching. The Region of Interest (ROI) is immediately identified using the area of matching feature points, which contributes to computational complexity reduction. Experimental results confirms that the proposed fingerprint recognition system achieves a tip-top recognition rate than that of the former competitive schemes on public fingerprint datasets (FVC 2000, FVC2002) and the dataset we collected.