Back to results list
Show full item record
Please use this identifier to cite or link to this item:
|Title:||New generation of automated fingerprint recognition system||Authors:||Liu, Feng||Degree:||Ph.D.||Issue Date:||2014||Abstract:||Fingerprint-based biometric is the most proven technique and has the largest market shares. It has been used for personal authentication for centuries and automated fingerprint recognition systems (AFRSs) have been used for decades. Although much progress has been made in AFRSs, the performance is still much lower than the expectations of people and theory estimation. Many new requirements are also raised along with more and more adoption of fingerprint technique in civilian applications, such as template security, hygiene, user-friendly and so on. For the purpose of further meeting people’s needs (e.g. recognition accuracy, template security, and hygiene etc.), this thesis explores two types of advanced AFRSs, namely high-resolution AFRS and Touchless 3D AFRS. For high-resolution AFRS, we firstly recommend an optimal reference resolution by theoretical analysis and experimental simulation based on two most representative fingerprint features, minutiae and pores. Such reference resolution is helpful to solve problems such as cost, interoperability, and performance of an AFRS, so as to benefits the establishment of optimal AFRSs. To improve the recognition accuracy based on features on high resolution fingerprint images, a novel hierarchical fingerprint matching method is then proposed. The approach directly matches features in fingerprints by adopting a coarse-to-fine strategy. In the coarse matching step, a tangent distance and sparse representation-based matching method (denoted as TD-Sparse) is put forward. In the fine matching step, false correspondences are further excluded by a weighted RANdom SAmple Consensus (WRANSAC) algorithm in which the weights of correspondences are determined based on their dis-similarity. High recognition accuracy is achieved since our proposed method is robust to noise and distortions of captured fingerprints and the inaccurate of extracted features. For touchless 3D AFRS, we firstly designed a touchless multi-view fingerprint acquisition device by optimizing parameters regarding the captured fingerprint image quality and device size. Optimization design of our device is demonstrated by introducing our design procedure and comparing with current touchless multi-view fingerprint acquisition devices. The efficiency of our device is further proved by comparing recognition accuracy between mosaicked images obtained by our proposed method and touch based fingerprint images. Then, 3D fingerprint images are generated by the proposed 3D reconstruction technique from captured touchless multi-view fingerprint images. The proposed reconstruction method puts emphasis on the correspondence establishment from 2D touchless fingerprint images and finger shape model estimation. Several popular used features, such as scale invariant feature transformation (SIFT) feature, ridge feature and minutiae, are considered for correspondence establishment. Binary quadratic function is found to be more suitable for finger shape model compared with another mixed model we proposed by analyzing 440 3D point cloud finger data collected by the structured light illumination (SLI) method. 3D fingerprint reconstruction results from different fingerprint feature correspondences are then given and the reconstruction accuracy is finally analyzed and compared. After that, 3D fingerprint features and their applications for personal authentication are studied. We define the 3D finger structural features, such as curve-skeleton, overall maximum curvatures as Curvature Fingerprint Features and investigate their distinctiveness for user authentication. These features are also used to assist fingerprint matching and make contribution to fingerprint recognition by combining with 2D fingerprint features. Since more information can be captured by touchless imaging, we propose an end to end solution for user authentication based on images captured by our designed touchless fingerprint acquisition device. Preprocessing steps including region of interest (ROI) extraction and image correction are implemented on the three views of raw fingerprint images captured by our device. New feature--Distal Interphalangeal Crease (DIP) based feature is then extracted and matched to recognize the human’s identity in which part selection is introduced to improve matching efficiency. Experimental results show the effectiveness of combining DIP-based feature with other features for touchless fingerprint recognition systems.||Subjects:||Fingerprints -- Identification -- Data processing.
High resolution imaging.
Hong Kong Polytechnic University -- Dissertations
|Pages:||xvi, 164 p. : col. ill. ; 30 cm.|
|Appears in Collections:||Thesis|
View full-text via https://theses.lib.polyu.edu.hk/handle/200/7430
Citations as of Oct 1, 2023
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.