Detection, Inspection, and Enforcement
3. Biometric Recognition
The Detection, Inspection, and Enforcement program is developing evaluation standards for advanced human recognition systems that will allow qualitative performance testing of recognition products over a wide range of watch-list and access-control applications. These systems presently include facial and fingerprint recognition.
3.1 Facial Recognition
To design methods for the evaluation of systems and technologies that identify humans in imagery and promote and advance face recognition technology designed to support existing face recognition efforts in the U.S. Government. Questions to be addressed include: has face recognition technology performance improved by an order of magnitude since the last formal independent evaluation; how well does it perform based on different image/video quality levels? Is there correlation with other biometrics such as iris? This technology will improve public services and benefit law enforcement and homeland security applications.
In prior fiscal years, NIST has been funded to conduct data gathering efforts, which have resulted in a human identification database. Using the database, NIST has developed the measurement methods for analyzing performance of facial recognition systems. The methods have proven sufficiently generic such that iris recognition systems can be evaluated in much the same manner. Indeed the measurement techniques and performance statistics are applicable to a wide range of biometrics, and non-human recognition tasks.
This effort is funded by the National Institute of Justice. Several image databases have been used for the NIST face tests. First, a NIST-gathered collection of volunteer images constitutes a database of multi-modal, multi-site, and multi-biometric collection. Second, 6.8M operational images from the Department of State give very large-scale performance numbers that are predictive of deployed performance in visa processing applications. Two additional databases collected at other government and academic facilities contain multi-modal images and video streams that can be used to test correlation between various biometric applications.
This effort will lead to increased efficiency and accuracy of verification/identification systems; allow quantitative performance testing of human recognition technology over a wide range of watch-list and access-control applications.
This project represents a continuation of the face and iris recognition evaluation projects that are anticipated to continue for at least another year. Emphasis will be on analyzing current evaluation results, measuring performance on various quality levels, and conducting correlation testing with other biometric modalities. An additional emphasis on performance from high and low quality video is anticipated.
These independent and unbiased technical evaluations are activities to which NIST is uniquely qualified. Related activities at NIST include face recognition technology development, evaluation of iris recognition, and mandated certification of government face recognition systems under the US Patriot Act.
NIST will conduct a series of independent, government evaluations on face and iris recognition and analyze the results to measure improved performance accuracy and correlation between biometric modalities. NIST will administer the Facial Recognition Verification Test (FRVT) 2006 and the Iris Challenge Evaluation (ICE) 2006 independent Government evaluations and write the final report for the Face Recognition Grand Challenge (FRGC).
NIST will administer the FRVT 2006 and the ICE 2006 independent Government evaluations, analyze the results and write the final reports.
A summary report will be written upon the completion of each biometric evaluation. The final report on the FRGC is to be completed by the end of the fourth quarter.
3.2 Fingerprint Fast Capture Evaluation
To evaluate fast capture devices for collecting rolled-equivalent fingerprints being developed under grants awarded by NIJ. Questions to be addressed include: does the device effectively capture fingerprints; with what speed; at what resolution, with what quality; and are the fingerprints captured compatible with legacy fingerprint systems and repositories? These devices promise an order of magnitude decrease in capture time, making it more efficient to collect rolled fingerprints from subjects. This technology, if proven, will improve public services and benefit law enforcement and homeland security applications.
Tenprint (all ten prints from the digits of the hands) capture technology today on average requires (under ideal circumstances) around 3 minutes to process a subject. New fast capture technologies promise complete tenprint capture on the order of 10 seconds.
Legacy fingerprints and systems have been designed to process inked fingerprint cards that have been scanned and live scan fingerprint images. To date, hundreds of millions of prints have been collected and enrolled into systems such as the Federal Bureau of Investigation's Integrated Automated Fingerprint Identification System (IAFIS) and the Department of Homeland Security's Automated Biometric Identification System (IDENT). The sensor technologies being applied in this fast capture initiative are considerably different. These include contactless sensors and photographic capture of friction skin detail. Tests are required to determine the quality of fingerprint images produced by these new devices and their compatibility with legacy fingerprint systems and repositories.
This project represents a new activity that is anticipated to go on for at least two years. These NIST evaluations are commencing at the end of the first year of NIJ funding to grant awardees, and the evaluations are expected to continue through at least the end of a second year of NIJ funding.
These independent and unbiased technical evaluations are activities to which NIST is uniquely qualified. Related activities at NIST include mandated certification of government fingerprint systems under the US Patriot Act, and mandated testing and certification for personal identification verification (PIV), which include the development and standardization of a fingerprint image quality algorithm along with software development kit (SDK) fingerprint matcher tests. This effort will lead to increased efficiency and accuracy; increased latent search ability improving prosecutions and forensic investigations.
NIST will receive prototype devices on loan from NIJ and conduct a series of investigative tests to evaluate the fingerprint images produced by these devices. Factors to be studied include: fingerprint image quality, fingerprint compatibility, capture speed, and device usability. It should be noted that this project will not include any large-scale data collection.
A summary report will be written and given to NIJ upon the completion of each device evaluation. Quarterly reports and an annual report will be provided.