Quality determination of fingerprint in travel document

National Authority for Identity Data

Algorithm that supports applications for travel documents and Dutch identity cards. It assesses whether a fingerprint taken is of sufficient quality for inclusion in the travel document.

Last change on 14th of June 2024, at 6:49 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: Field not filled in.
• Status: In use

General information

Theme

Organisation and business operations

Begin date

Field not filled in.

Contact information

[e-mail]

Link to publication website

 

Responsible use

Goal and impact

Used as technical support to assess whether a fingerprint taken is of sufficient quality

Considerations

Measuring fingerprint quality increases the reliability and accuracy of checking identity (identity verification). It also facilitates cooperation between different fingerprint recognition systems. This ensures fewer failed comparisons of live fingerprints with the fingerprint recorded in the document.

Human intervention

If the index finger is injured or missing, we will take a fingerprint from another finger (the middle finger, ring finger or thumb). If it is impossible to take two fingerprints, we may ask for a statement from a qualified doctor or medical institution. If we cannot take fingerprints, we will state in the application whether this is permanent or temporary.

Risk management

No basic security test was taken

Legal basis

Under Section 3(2) of the Passport Act, there is an obligation to include the fingerprint in the travel document. This concerns a requirement set by the European Union (Regulation 2252/2005). Article 28a of the Netherlands Passport Implementation Regulations 2001 describes the manner in which the fingerprints of the applicant for a travel document are recorded and in which cases this may be waived. The quality of the recorded fingerprint(s) is decisive in this (paragraph 2-3)*. The aforementioned provisions provide the legal basis and thus lawful processing of the fingerprints as referred to in Article 6, third paragraph, of the AVG in conjunction with Article 9, second paragraph, under g, of the AVG. *See also Article 40a of the Passport Implementation Regulations Caribbean Countries and Article 42a of the Passport Implementation Regulations Abroad 2001.

Operations

Data

- NFIQ2 source code and user guide; - NFIQ2 conformance test; - ISO Biometric Sample Quality; - NIST Interagency Report 8382; - Passport Act; - Passport Implementation Regulations The Netherlands 2001; - Passport Implementation Regulations Abroad 2001; - Passport Implementation Regulations Caribbean Countries The training set consists of 6,629 images (3,295 in class 0 and 3,334 in class 1) carefully selected from the AZLA, POEBVA and DHS2 datasets. The selection rule was as follows: - Class 1 consists of images with NFIQ 1.0 value 1 (with activation score > 0.7) and true score in the 90th percentile for each of the NFIQ 2 providers. - Class 0 consists of images with an NFIQ 1.0 value of 5 (with activation score > 0.9) and a true score less than a threshold corresponding to a false match rate of 1 in 10,000, i.e. false rejection at a false match rate of 0.0001. Furthermore, 99.797 images were randomly selected for model validation. The data used in the development of NFIQ 2 came from the Federal Bureau of Investigation (FBI), the Department of Homeland Security (DHS), Los Angeles County Sheriff's Department (LACNTY), Arizona Department of Public Safety (AZDPS). The Bundeskriminalamt (BKA) in Germany tested NFIQ 2 on its data. All images were 8-bit greyscale. The images were pre-compressed using Wavelet Scalar Quantization (WSQ) compression.

Technical design

NFIQ2 software reads a raw or WSQ-compressed fingerprint image, calculates a set of image quality characteristics and uses these characteristics to predict the usability of the images. We make the prediction using a random forest classifier trained using comparison scores from different commercial fingerprint comparison algorithms from different operational fingerprint databases. NFIQ 2 software produces a quality score consistent with the international fingerprint quality standard ISO/IEC 29794-1:2016 is in (0-100), where 0 means no utility value and 100 is the highest utility value. NFIQ 2 uses a random forest model to classify the data. NFIQ 2 is a binary classifier. Scores are derived from the random forest decision. Each decision tree uses a random subset of the feature vector. When using 100 decision trees, each vote is one score point. The distribution of votes/score points is non-linear, and not necessarily uncorrelated. NFIQ 2 uses 69 features manually selected and validated.

External provider

Identity/Services