In real-life scenarios, fingerprints are not usually acquired in perfect conditions. Fingertips could be dirty or wet and affect the quality of fingerprints sensed by a fingerprint sensor. Fingerprints could also be acquired in rotated forms. When the humidity is high, fingertips are moist due to sweat, and the acquired fingerprints could result in low performance of devices or systems that utilize fingerprint sensors.
BVCUNN-Index-fingerprint dataset comprises a total of 1400 fingerprints of 200 subjects consistently acquired at different conditions which include: various levels of rotations, dirty and moist conditions. The rotations correspond to angles 0º, 10º, 20º and 30º constituting four conditions. The dirty and moist forms are the fifth and sixth conditions.
Images of fingerprints were acquired using a Microsoft fingerprint reader and stored in Bitmap (BMP) format. This database was collected in single-session in 2 days in the Department of Electronic Engineering, University of Nigeria. the size of this database is 97.8MB (zipped).
Two impressions per subject were acquired for the normal condition, 0 , while one impression per subject was acquired for other five conditions.
A user’s fingerprint is placed on the scanner following these rotations. In order to stabilize the fingerprint scanner and follow accurate rotations during enrolment a comfortable method was improvised. A porous Polyvinyl Chloride (PVC) sheet was cut to a dimension of 20 cm by 15 cm and angular displacements of 10°, 20° and 30° were constructed to the right and left directions from the normal, 0 . The shape of the fingerprint reader was marked and carved on the PVC material and the fingerprint reader was fitted and secured with masking tape.
Each user placed his finger on the scanner, twice for normal orientation, 0 , and once for degrees: 10°, 20° and 30°.
In order to acquire dirty fingerprints subjects touched their fingers tips on a surface of dust before scanning. In order to acquire moistened or wet fingerprints, subjects moistened their fingers with water prior to scanning.
Fingerprints are numbered following this nomenclature Fnnn_C.bmp where:
F is a prefix for finger
nnn represents the subject number or identifier and runs from 001 to 200.
C stands for the conditions of acquiring a subject’s image and could be:
A0 – symbolizing normal condition 1
A01 – symbolizing normal condition 2
A10d or A10D – symbolizing 10 degrees’ rotation
A20d or A20D – symbolizing 20 degrees’ rotation
A30d or A30D – symbolizing 30 degrees’ rotation
D – symbolizing dusty fingerprints
M – symbolizing moist fingerprints
The nomenclature for this database can be assessed from the readme file (readme_indexprints.txt)
When the BVCUNN-Index-Fingerprint dataset is used in any form of research, please the following paper should be cited:
Ogechukwu N. Iloanusi and Celestine A. Ezema,
A quantitative impact of fingerprint distortion on recognition performance,
Information Security Journal: A Global Perspective,
Keywords: Fingerprint, security, biometrics, image quality, distortion, performance, regression analysis