PHLburg Technologies, Inc.

Information Sciences

Automatic Fingerprint Matching

The Customer’s Printrak Biometrics R&D group is pursuing and seeking critical analytical technologies to improve the accuracy and efficiency of fingerprint matching systems for a wide range of public safety applications.  Timely insertion of such technologies into the Omnitrak system and its eventual integration into the Customer’s seamless communication medium will significantly improve our posture in the world marketplace

Problem Background:
As a consequence of recent events and real concerns for the future well being of the free world, public safety has become the centerpiece of our global society.  The Customer’s relevant expertise in fingerprint matching, based on a 30-year history of successful development and deployment of such systems worldwide, constitutes one element of the larger public safety enterprise. Work is already under way to extend this technology to include other biometric and investigative information sources, and to deliver it not only to the desktop but also wirelessly through the Customers’s broad spectrum of communications systems. While this evolution is inevitable, the demand for accurate and rapid identification of fingerprints is rising.  We are in need of analytical techniques that can deal effectively and efficiently with the underlying disciplines of image processing, feature extraction, database indexing and matching involving imperfect real-world fingerprint data.  While we are diligently addressing some of these issues, we believe that external expertise could accelerate our efforts and allow us make timely advances into the larger integrated, information rich public safety systems.

Specific Technical Needs:
Included among our most immediate needs are the following with emphasis on accuracy, speed and automation.

1.    Image Processing
a.    Segmentation of multiple fingerprints from a composite image.
b.    Isolation of relevant image from the background.
c.    Separation of overlapping fingerprint images.
d.    Image enhancement.
e.    Quality and distortion maps

2.    Feature Extraction
a.    Minutiae detection.
b.    Singularity detection.
c.    Local ridge structure
d.    Image orientation.
e.    Classification

3.    Filtering and Matching
a.    Feature filtering
b.    Feature matching
c.    Minutiae matching
d.    Local structure matching.
e.    Latent, 10/20-print matching
f.    Distortion compensation
g.    Fusion

4.    Reverse Search and Investigative Reasoning
a.    Identity search
b.    Cold case history
c.    Matching pattern deduction
d.    Crime pattern discovery

It should be mentioned that the Customer is currently using the disciplines underlying most of these items quite successfully. However, about 30% of real-world crime prints encountered defy successful resolution because of poor image quality, interfering background, overlapping, distortion, etc. Of these, at least 50% could be resolved visually.  The Customer believes that critical extensions of the basic techniques as well as other new image processing and deductive methods could help accomplish automatically much of what a human can discern visually.  This level of improvement would put the Customer on par if not ahead of its competition. Incorporating this technology into a larger biometric subsystem within the Customer's current and planned information and communications space would make the Customer practically unrivaled.

To view this Problem Statement in a PDF format, click here.