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

Abstract

In order to assess longitudinal effects in fingerprint biometrics several studies have been carried out in the last decade. Almost all of these investigations focused on non-public forensic fingerprint datasets because there is hardly any publicly available fingerprint database allowing experiments and a corresponding analysis of longitudinal aspects is often infeasible as the contained fingerprint samples are either captured in one session only or with a short time interval of only a few weeks between the sessions. With this work a new fingerprint dataset (108,106 samples) is introduced which can be used for inter-session (longitudinal) as well as inter-sensor fingerprint investigations. A total of ten different capturing devices based on distinct sensing technologies (optical, capacative, thermal, multispectral) is utilized and the imprints of all ten fingers of each of the 50 participating volunteers were acquired at four different time-separated sessions over 2 years. This enables inter-session as well as cross-device evaluations. Additionally, a first quality and recognition performance evaluation deploying several state-of-the-art minutiae based fingerprint recognition schemes and NFIQ 2.0 as quality measure is presented.