PoreNet: CNN-based Pore Descriptor for High-resolution Fingerprint Recognition

Abstract

With the development of high-resolution fingerprint scanners, high-resolution fingerprint-based biometric recognition has received increasing attention in recent years. This letter presents a pore feature-based approach for biometric recognition. Our approach employs a convolutional neural network (CNN) model, DeepResPore, to detect pores in the input fingerprint image. Thereafter, a CNN-based descriptor is computed for a patch around each detected pore. Specifically, we have designed a residual learning-based CNN, referred to as PoreNet that learns distinctive feature representation from pore patches. For verification, the match score is generated by comparing pore descriptors obtained from a pair of fingerprint images in bidirectional manner using the Euclidean distance. The proposed approach for high-resolution fingerprint recognition achieves 2.56% and 0.57% equal error rates (EERs) on partial (DBI) and complete (DBII) fingerprints of the benchmark PolyU HRF dataset. Most importantly, it achieves lower FMR1000 and FMR10000 values than the current state-of-the-art approach on both the datasets.

Interpol review of fingermarks and other body impressions (2019 – 2022)

“…At the border between detection and identification, we note the use of a Reflected Ultra Violet Imaging System (RUVIS) to systematically capture marks on non-porous surface and search them against biometric records taking advantage Convolutional Neural Networks (CNN) algorithms based on level 3 features [ 45 ]. CNN are undoubtedly the new architectures to successfully deal with the extraction and matching of level 3 (more specifically pores) features [ [46] , [47] , [48] , [49] , [50] , [51] , [52] , [53] ]. Combining level 2 and level 3 features improves the latent recognition rate in comparison to the minutiae matching [ 54 ].…”

The PLUS Multi-Sensor and Longitudinal Fingerprint Dataset: An Initial Quality and Performance Evaluation

“…In case of minutiae-based applications examples for CNN approaches have been introduced and discussed in [22], [36], [42], [54], while examples of commercial state-ofthe-art recognition systems are VeriFinger SDK and Innovatrics 2. https://github.com/usnistgov/NFIQ2 ANSI 3 . In case of non-minutiae-based applications CNN-based methods have been investigated in [1], [4], [64], while approaches utilising traditional methodologies like FingerCode or Phase-Only-Correlation have been introduced and discussed in [48], [50] and [30], [31], respectively. FingerCode and Phase-Only-Correlation are designed to have a high stability against certain distortions, especially for poor quality samples which can often be collected at crime scenes.…”

“…All the listed systems can be applied to the proposed multi-sensor and longitudinal dataset and this list provides suggestions for further systems that can be employed to investigate the dataset. In this study the focus is on the publicly available minutiaebased NIST Biometric Image Software (NBIS) 4 application and two commercial state-of-the-art minutiae-based ones: ANSI & ISO SDK developed by Innovatrics and VeriFinger SDK 11.0 developed by Neurotechnology. NBIS was selected as it has been used in previous studies on FP ageing effects, e.g.…”

Multiresolution synthetic fingerprint generation

“…With the development of high-resolution sensors that are able to capture L3 fingerprint images, researchers saw an opportunity to devise more accurate recognition approaches by using extra information, such as sweat pores [3][4][5]. Besides, L3-based approaches improve security by hindering spoof attempts [6][7][8].…”