Papillon Systems

PAPILLON: UP-TO-DATE BIOMETRIC SOLUTIONS
from Russia

  
INFORMATION PRESS-OFFICE PRODUCTS PERFORMANCE SITE MAP

Q     m     [x]  

    Sign up   
 РУС | ENG | ESP 
AFIS Selection Guidance
 

The AFIS’s efficiency is generally measured by two interrelated abilities – reliability and accuracy of search.

Reliability is a percentage of true mates selected by the AFIS per their total number in the database.

Accuracy (selectivity) is a percentage of pairs falsely defined by the AFIS as true mates per total number of comparisons.

Yet another significant characteristic of any AFIS is its performance – a number of comparisons per time unit. At equal showings of reliability and accuracy, a system with low performance appears to be more costly in its maintenance, since it requires more matchers for operation, more powerful equipment, and more charges for installation and servicing.

Quantitative values of reliability and accuracy are not constants for one and the same AFIS since they feature search results against a specific tenprint and latent data array. And so, not every system demonstrating high showings during test running is able to prove the same results while operating with real fingerprint data files containing hundreds of thousands, or even millions and tens of millions, of tenprints.

In a number of cases, the PAPILLON AFIS substituted for some AFISs of other vendors (ссылка на протоколы) installed in some of the MOI divisions. The same array of paper tenprints and latents was converted to the PAPILLON AFIS database as that the predecessors operated with. As a result of the PAPILLON AFIS search, hundreds of subjects were else identified though not recognized by the predecessors as true matches.

For more details…

It should be noted that the substituted AFISs have also high values of search reliability and accuracy as declared by their vendors and proved at testing many times.

The above-said allows concluding that once obtained high values of search reliability and accuracy cannot serve as a criterion for evaluating the AFIS’s quality and efficiency. A sure proof of the system efficiency is constancy of its searching capabilities as applied to databases of any size and its tolerance to any distortions of both objective (qualitative differences between fingerprint data arrays) and subjective (human) nature.

An AFIS that may truly be called effective is a system providing high selectivity of search, showing stable and steady results on databases of any size and any fingerprint record quality and producing candidate lists that need minimum time for an expert to verify them and to make the final hit-no-hit decision.

Stable and steady searching capabilities of an AFIS depends upon quite a number of factors: ability to operate equally reliably with fingerprint images of different quality, method of pattern description and selectivity of comparison algorithms, perfection of recognition and pattern coding algorithms, degree of automation of all stages of tenprint and latent processing.

Let us dwell upon each of these factors:

  1. The higher is the quality of tenprints and latents entered into the AFIS, the more informative they are and more accurately the system matcher works providing high search selectivity and minimizing the candidate list length. Unfortunately, considerable part of actual paper tenprint files and a vast majority of latent prints lifted at crime scenes do not fall into the category of high-quality source information. Of course, it might be possible to exclude the influence of the factor and not to enter into the AFIS tenprints and latents of poor quality at all. However, in practice, such an approach tends to result in reliability degradation by a percent of rejected objects, and thus, it is considered unacceptable. The system has to be capable of working with complicated objects including those of unknown scale.

The PAPILLON AFIS imposes no special requirements to the quality of submitted tenprint cards and latents as compared to manual processing. Unknown scale of a latent image, age-related changes of fingerprints, pattern distortions over any skin strain are regarded in the recognition algorithms and have no impact on the reliability of identification. The system is capable of restoring a latent print distorted through overlay of another latent print or through a background texture.

Up-to-date technologies of data entry that are developed by PAPILLON contribute greatly to fingerprint image enhancement:

– Electronic fingerprinting (see PAPILLON LIVE SCANNER)

– Video input (through the PAPILLON video input system) and processing of latent prints using special filters of the PAPILLON RASTR system (separation of overlapping latents, ridge pattern amplification, subtraction of background texture, etc.)

  1. Method of mathematical description of ridge patterns and matching algorithm it defines are governing factors in providing search reliability and accuracy. Three groups of specific features constitute a description of any ridge pattern, namely:

Integral structure of ridge pattern - a set of directions formed by friction ridges. Focal points such as cores and deltas can be also used for description.

Friction ridge details (minutiae) – endings and bifurcations of ridge lines. Location and possibly angular orientation (direction of ridge lines at minutiae points) of minutiae are accounted.

Ridge count and relations (topological characteristics of a ridge pattern) represent the relative position of neighboring minutiae across and along the ridge flow. These data are the most powerful criteria for comparing ridge patterns.

Actual AFISs uses either one of the above-listed characteristics or a combination of them. The more complete and in greater detail is the pattern description, the higher is the search performance of an AFIS provided that its image recognition algorithms and automatic coder are capable of extracting the specified properties with required accuracy. Thus, comprehensiveness of a realized pattern description method is a measure of how the developers of a given AFIS have succeeded in resolving the problem of automatic recognition and coding of images.

In the PAPILLON AFIS, all three groups of characteristic features are used for most comprehensive description of ridge patterns. Topological approach to pattern description and hierarchic method of their comparison provide high selectivity at searches which surpasses the one of those systems that describe only an integral structure and (or) position and direction of minutiae.

  1. The more thorough are automatic algorithms for image recognition used in the AFIS, the more reliable is the mathematical model used for comparison and the less is the operator’s interference and influence on the process of coding.

Choosing the most informative method of pattern description in the PAPILLON AFIS/APIS imposes stronger requirements on the automatic coder of the system with which it copes well. Fingerprints available on tenprints are coded automatically. Coding latent prints is a semiautomatic process, the portion of operator’s actions depending upon the latent image quality.

It should be noted more particularly that the capability of operating with palmprints including latent palmprints is among the main AFIS’s merits. It is known to specialists that 15-30% of latents lifted at crime scenes are referred to palmprints. As known from the experience of law-enforcement agencies using a palmprint version of the PAPILLON AFIS, the efficiency of automated fingerprint and palmprint files is 20-35% higher as against a fingerprint version of the AFIS. Please note that coding palmprints in the PAPILLON AFIS/APIS is completely automatic.

Thus, to estimate the efficiency of different AFISs, one should take into account the following aspects:

  • Comprehensiveness of the method realized for mathematical description of ridge patterns
  • Accuracy of the automatic coder, which can be indirectly assessed by amount of operator’s involvement into coding processes
  • System performance
  • Size of candidate lists produced by the system
  • Position of true mates on candidate lists
  • Ability to operate with real tenprints and latents without preliminary quality selection
  • Ability to operate with non-uniformly scaled images
  • Ability to operate with palmprints including latents
  • Use of up-to-date high-quality techniques of data entry (electronic fingerprinting, live video input)
  • Implemented projects, their range and scale (size of databases in operation)

Objective and official comparative assessment of automated fingerprint systems can be obtained only through a full-scale operational testing of two or more AFISs against real fingerprint and palmprint data arrays using elaborate experimental procedures having provided absolutely equal conditions for each system under test.

  Copyright 2001-2018 Papillon Systems. All rights reserved.