Automated Fingerprint Identification
Goals of Module
Understanding of the basic concept
of the AFIS
Arrest/Applicant Fingerprint and
Palm Card Processing in AFIS
Two Finger vs. Tenprint File
Tenprint vs. Tenprint File
Tenprint vs. Unsolved Latent File
Expert Matching currently used in
the state system
Impact of agency workflow design
Current Issue’s/Limitations of AFIS
Goals of Module
Brief description of a latent crime scene
Latent Acquisition in AFIS.
Latent Finger/Palm Searches
Latent vs. Latent Cognizant File/Latent vs. Unsolved
File/Palm print vs. Unsolved Latent Palm
Latent Palm vs. Palm print Filet/
Latent Palm vs. Unsolved Latent PalmTP
Accuracy and how it is established in a
Differences in the state and FBI latent
Current Issue’s/Limitations of latent
Automation of Fingerprint Search
FBI / National Bureau of Standards Begin a Long Term Collaboration to Automate
1963 Fingerprint Search Process
R&D Intentionally in the Public Domain
1979 FBI Introduces Initial Automated Processing
Commercial AFIS Systems Emerge and are Rapidly Adopted
1980’s Unable to Exchange Fingerprints Between Vendors
FBI Processing Times and Backlogs Fall Behind States
1986 ANSI/NBS-ICST 1-1986 Fingerprint Identification - Data Format for Information
1989 FBI and Advisory Board Conceive National Search Hierarchy
And Plan to Revitalize FBI Identification Services
1992 Advisory Policy Board Approves FBI IAFIS System requirements.
Goals of Automating the Fingerprint
Automation of the Fingerprint Search Process
Ability to effectively Search Crime Scene
•Cost Savings (no files or storage required )
•Ability to search Crime scene prints
•Improve quality of prints captured
FBI Files 1997
AMERICAN INSTITUTE OF
“The basis of fingerprint identification is the
premise that the configurations formed by the
raised ridges of the palmar surface of the
hands are unique and do not undergo any
natural changes, except growth, from fetal life
a.k.a. Minutiae or Points
Specific traits found in friction ridges used to establish an
identification by their relative location to each other.
The Average finger has between 75 and 175 points of
identification. Palmar area approximately to 2,000 points of
e.g. ending ridges, dots, bifurcations, islands
Core Systems in all
Automated Fingerprint Identification Systems
Input Coder Matchers Database
No matter what size the system, a small SPEX system or the FBI, they all have these
Latent Input from Lincoln PD, Omaha PD, Douglas Co. and Stat Patrol
Input of Latent Print
The examiner will enter the
latent examiner will enter the
latent lift/photograph into the
system. The image is
designed for the fingerprint
to be captured with the
camera. The camera
produces the better images.
The scanner captures a
larger area and is better
suited for a latent palm. The
scanner samples the image
at intervals to create the
image so using the scanner
you do have a slight loss in
The Search (Coders & Matchers)
The examiner has a choice to manually select the points of identification (minutiae) or allow
the system to select it.
Search database is approximately 350,000 records in Ne. and 60 million at the FBI.
Other helpful parameters that can be set are age, classification and gender.
Search takes 4-13 minutes depending on search parameters in Ne.
All 20 images rolled & flats are searched, if available.
Current system is 85% accurate when the latent has 15 points of identification and the
target (exemplar) ten print card has the corresponding 15 points.
Avg. Latent has between 10 & 30 points.
Approximately 30% of crime scenes produce useable latent images. 21% of the crime
scenes have latent fingerprints. The Other 9% are latent palm images
The system uses expert matching to develop a list of viable candidates for the examiner to
Accuracy ranges from 85% with the target print in the 1st. 5 candidates to 88% that the target is in
the top 15 candidates.
System creates a candidate list of the most likely suspects and the latent examiner determines hit
or no hit.
If the lift is not identified it can be added to the unsolved latent file.
Latent Palm Print Searching
The new systems now have the ability to
search latent palm images.
30% of the useable latent prints developed at
a crime scene are palms.
Palm system accuracy is 70 to 90 % if the
latent has 15 points of identification and the
target (exemplar) ten print card has the
corresponding 15 points.
Palms have in approximately 2,000 points of
Palms have dedicated matchers and coders.
Application software supports 360 degrees
Lower palm (thenar, hypo-thenar and
interdigital) Upper palm and Writers Latent
palm searching will use a distribution point,
segment, and hand areas to improve
Unsolved Latent Files
(latent finger & latent palm)
The latent finger and palm print files in Ne. are sized to hold 75,000 latent
records including both latent fingerprints and latent palm prints.
Regardless of whether the record was entered from a ten print inked card or
submitted via LSS, workflow capabilities are provided for automated search of
the unsolved latent file.
When a new fingerprint card is processed through the ten print system and it is
not identified the minutiae are compared with the unidentified latent stored in the
UL file on AFIS.
Latent finger against the Unsolved accuracy 88-93%, a palm is 73-93%.
Accuracy figures used in this presentation are from a controlled environment and
are probably significantly higher than the results obtained in every day searching.
There are a number of factors that will influence accuracy:
Quality of the ten print card.
Quality of database , current database contains images captured at 500 ppi and
Quality of the examiner
Universal Latent Workstation
If a latent is serious enough it can be
searched against the FBI IAFIS system.
FBI system consists of :
The image captured for the state afis
search can be used for the FBI search but
new minutiae needs to be plotted.
50+ million arrest fingerprint cards from
every state in the union at 500ppi.
Database penetration less than 30 %
Search database is created from the 1st.
Card received. The minutiae however is
updated when subsequent cards are
processed, so they have a composite of
the best minutiae.
FBI has an unsolved latent file.
Core IAFIS Service
Search Database Make Up
Copy of all/specific # of cards processed
Copy of the 1st. Card processed (composite of all minutia
Composite of the best fingerprint images.
System accuracy is influenced by:
Quality of images in the database
Proficiency testing/ remediation training
The FBI accuracy is 95 to 98%. The database
has the best quality images of any database in
the country. I would estimate the other state
and local databases around 90%
Lives can Unit
Today over 70% of the
fingerprint cards are
captured and digitized
by lives can units.
Inked cards are the
accepted at state level.
Images are captured at
either 500 or 1,000 ppi
Good quality finger/palm images
are critical to the AFIS database.
Prints sent from the coder are
viewed by a trained technician.
Thresholds monitored by the
system administrator who deter-
mines the amount of work going to
The current system is designed for
5-10% of the transactions to go to
Advanced Fingerprint Processor (Coder)
1)The original captured fingerprint
2) The enhanced image is then “thinned” so that a “skeleton” image
3) Individual minutiae are then extracted from the thinned image
4) minutiae processing algorithm contains special procedures that
minimize the number of “false” minutiae arising from over-inking,
under inking, or smudges.
The newer systems are now using the 8 best rolled images for
the ten print search. Older systems oft used 2 fingers, NYC
used the thumbs while NYS used the index fingers. The index
and thumbs usually have large numbers of minutia so this
was a pretty accurate system when storage costs were more
Lights Out Functionality
This allows the system to make a determination
of Identification or Non-Identification. Currently
NYS is using this on non-criminal transactions
with excellent results. It could be a way for a
small state system like Nebraska to supplement
ten print examiners. Us the lights out for the
initial verification. The system administrator
would be required to monitor thresholds more
closely but it would improve system accuracy
without increasing staffing requirements.
Ten print Results
The results of the ten print search are the basis
Create criminal history (state rap, NCIC etc.)
Insure no criminal history (applicant record check)
An error in either of the above can dramatically
impact public safety.
All live scan sites in Nebraska are required to capture a
mugshot at printing. This will allow the state to move into
Facial recognition in the future. In the current system
mugshots are available to any latent workstation
or live scan unit. This for example would allow Lincoln to get a mugshot
from Scotts Bluff on a current arrest.
Currently not all sits are following NIST guidelines for taking mugshot
photographs. In the future all agencies will b required to follow the guidelines
to share photo’s as well as creating a facial recognition database.
DOE, JANE HENRYETTA
PHOTOS OF SMT’S
USE SAME CAPTURE TECHNIQUES AS
CAPTURE VISIBLE SMT AND THOSE
DISCLOSED DURING BOOKING
DOE, JANE HENRYETTA
HUMAN FORMS AND FEATURES
ANIMALS AND ANIMAL FEATURES
REMOVED & SYMBOLS
ARM , LEFT
ARM , RIGHT
FOOT , LEFT
FOOT , RIGHT
HAND , LEFT
HAND , RIGHT
HEAD , NON SPECIFIC
Allows the Livescan operator to
verify an identity using
The system uses a form of lights
out and if the person has a
fingerprint record it can confirm
Returns information and a
Increases Officer safety :
Obtains an accurate record of a subject in
Identify potentially dangerous/wanted
individuals in the field.
Utilizes multi-biometrics for increased
Operates using cellular phone technology.
Lights out Operation, fingerprint expert not
Mobile Units which can work anywhere there
is cellular availability
Remote Single Finger Identification Pilot