PAPILLON AFIS Multibiometric Identification System
On offer: AFIS-8.4 and AFIS-9
Multibiometric data banks of unlimited size for such criminal justice and civil purposes as personal registration/identification:
– by fingerprints and palmprints
– by facial images
– by iris images
Automatic processing and identification of incoming biometric information, including fingerprints and palmprints from crime scenes, for forensic experts, law enforcement and supervisory authorities, other government agencies.
Real-time biometric instant identification of individuals.
PAPILLON AFIS supports import/export of information (finger and palm print images, appearance and iris images, text data) for registration and/or database checks.
Replenishment of PAPILLON AFIS databases is possible through desktop-clients, from other databases, and with the use of the following PAPILLON products:
Reliability and accuracy of fingerprint and palmprint searches (50 million tenprints in the database)
Average size of candidate lists:
Tenprint to tenprint: 1-2 candidates, the true candidate appears at the top in 100% of the cases with 100% hit rate
Latent to tenprint: the length of the list is 20 candidates, the true candidate appears at the top in 85% of the cases, in 90-95% of the cases the true candidate is among the first five entries of the candidate list with a hit rate of at least 85%
Tenprint to latent: 2-3 candidates, in the vast majority of cases the true candidate is at the top of the list with a hit rate of at least 85%
Import/export formats:
Papillon
ANSI/NIST (RUS-I, Interpol, EFTS, EBTS)
Conformity:
FBI WSQ compression of images (compression of finger and palm print images without noticeable loss of quality; the maximum compression rate is 1:15)
FBI “IAFIS IQS”: CJIS-TD-0110 and CJIS-RS-0010 (quality of images taken with fingerprint scanners meets the FBI specifications)
Dactyloscopy GOST R 58298-2018, ISO/IEC 19794-2 and 19794-4 (biometric data exchange format complies with the Russian Federation national standard)
Full-face image ISO/IEC 19794-5 (the format of biometric data exchange corresponds to the Russian Federation national standard)
Backup server – module for backing up critical database components
Data Exchange Server – module for converting information from other databases (Oracle, etc.)
Data repository – data storage and backup server. It consists of disk drives and an Oracle/PostgreSQL server designed to allocate the text part of the AFIS database as a separate database, open for work through web browsers and interoperation with external systems.
Subject to licensing:
Operator workstation (the number of simultaneous sessions is subject to licensing, the number of operators is not limited)
Local workstation (the number of simultaneous sessions and the possibility of remote access to the centralized AFIS database are subjects to licensing)
Matchers (restrictions apply to the number of searches for a specified period of time. E.g. the latent-tenprint search allows for 50 requests per day)
Central server
Not subject to licensing: Backup server, data exchange server and data repository.
high-quality photography of latent prints in different illumination modes (PAPILLON-Fosko, PAPILLON-ExpertLab)
AFIS tools and filters to improve the quality of latent prints, separate overlapped latent prints, remove underlying texture, etc.
High-precision proprietary algorithms for recognition and comparison of biometric images (fingerprints, face and iris images)
Work with latent and known palmprints
Identification of a latent print left by phalanxes of a ring finger in the flexor line area by a full palm print image
Work with latents containing a small number of minutiae:
fingerprints with at least four and palmprints with at least six minutiae are suitable for input into PAPILLON AFIS
in the Russian AFIS databases, about 43% of latents have a small (<17) number of minutiae, which amounts to ~29% of all cases of identifications
Identification of a poor quality latent by six minutiae
No special requirements to input fingerprint images in a uniform scale: compensation for any scale changes and distortions of fingerprints: due to age, post-mortem, resulting from trauma or disease.
Identification of a fingerprint with a strong deformation of the papillary pattern
Automatic encoding of all tenprints and latents:
high speed of arrays formation
elimination of encoding errors and associated omissions of “tenprint-to-tenprint”, “latent-to-tenprint”, “tenprint-to-latent” matches
High reliability and selectivity of comparisons, incl. the following cases:
latent and known palmprints
unknown position of a finger or palm
unknown type of finger pattern
any deformation or distortion of a papillary pattern
unknown orientation of the latent relative to the fingerprint
scars and skin lesions
a few minutiae on the latent
Stability of search performance on any size of the database without prior selection of tenprints and latents by quality
A special algorithm of work with tenprints of the unknown dead: the highest probability of matching with an life-time tenprint
Identification of a quasi-latent from the tenprint of the unknown dead that had been lying in a bog for a long time. The life-time tenprint was made 22 years earlier.
High precision “tenprint-to-latent” searches (most manufacturers of criminal justice AFISs did not solve this issue):
short candidate lists with the native candidate at the top positions
complete symmetry of “tenprint-to-latent” and “latent-to-tenprint” searches in all parameters:
reliability and accuracy
number of identifications and false matches per million comparisons
time spent on verifying candidate lists
Search symmetry: tenprint-latent
Search symmetry: latent-tenprint
Search against plain impressions:
Not only rolled fingerprints but also plain impressions are involved in latent-tenprint and tenprint-latent searches
About 11% of all matches for latent prints are those made against plain impressions, which would be impossible without this type of search (in data arrays, where the percent of tenprints containing slaps is no less than 50%)
Tenprint-tenprint search by palmprints: positive identifications even if the fingerprint quality in tenprint cards is very low
Tenprint-tenprint search by face and/or iris images:
improves the reliability of identifications in AFIS
allows information about personal identity to be obtained even in the absence of fingerprint information
expands the use of AFIS
Special methods (filters) for working with candidate lists generated for latents:
reduce time for verifying top (visible) parts of candidate lists to get identifications
enable additional identifications from invisible (hidden) parts of candidate lists that are not available if the filter is not applied
Neural network algorithms for analyzing candidate lists:
enable 80-90% identifications from top (visible) parts of candidate lists, reducing labor costs by tens of times
enable extra identifications from invisible (hidden)parts of candidate lists that are not available when the traditional method of verification is used
Technology of automatic express ID checks against tenprint databases in real time:
Requests from fixed and mobile stations (IP connection) are received to verify persons’ identity
Persons’ identity is established/confirmed by fingerprint and/or facial images and/or demographic data
Check time is a few seconds (if high-speed communication channels are used)
Result of the check contains text data and photos from all matches (tenprints) found in the database
Technology of automatic express checks of fingerprints against an array of unsolved latents:
Revealing the facts of the person’s complicity in illegal activities or crimes previously committed
Result of the check:
positive – the system is sure that the latent from unsolved crime scene found in the database belongs to the person being checked
confirmation required – the result must be reviewed and confirmed by a forensic expert
negative – no unsolved latents belonging to the person being checked were found in the database
Face recognition
Finger and palm print identification
Footprint identification
Iris recognition
