Systems for Fingerprint Classification and AFIS – Comprehensive Study Notes

Fingerprint Pattern Fundamentals

Three fundamental classes of fingerprint patterns: Arch, Loop, Whorl.
- Serve as the basis for every major classification system that followed.
Detailed subclasses (Galton’s original terminology maintained throughout):
- Arch → Plain Arch, Tented Arch.
- Loop → Ulnar Loop (opens toward little finger), Radial Loop (opens toward thumb) – remember handedness flips the designation.
- Whorl → Plain (Classical) Whorl, Central Pocket Loop, Double Loop Whorl, Accidental Whorl.
Relative frequency in general population (key for probability arguments & rarity-based sorting):
- Arches ≈ $5\%$ .
- Loops ≈ $60{-}65\%$ .
- Whorls ≈ $30{-}35\%$ .
Core: Approximate centre of a print; reference point for ridge counting & pattern location.
Delta: Triangular zone where ridge flows diverge; critical landmark for distinguishing loops/whorls from arches.

Arch Family

Plain Arch: Ridges enter one side, rise gently, exit other side; no delta; rarity makes them useful for quick elimination or prioritisation.
Tented Arch: Sharper up-thrust; may display a core and/or a delta despite being arch category.

Loop Family

Ulnar Loop: Ridge flow toward ulna (little finger side); has 1 core & 1 delta; at least one ridge re-curves between them (illustrated on left hand slide).
Radial Loop: Ridge flow toward radius (thumb side); mirror logic of ulnar; illustrated on right hand.

Whorl Family (Overview)

Plain Whorl (a.k.a. simple whorl): ≥1 ridge completes a full circuit; 2 deltas; at least one recurving ridge touches an imaginary line drawn between the deltas.
Central Pocket Loop: Also 2 deltas & a core; however, the imaginary delta-to-delta line does not touch the recurving ridge(s).
Double Loop Whorl: Two distinct loop formations (distinct shoulders) + 2 deltas.
Accidental Whorl: Composite of two different pattern types, ≥2 deltas; often extremely complex (three deltas illustrated).

Early Criminal Identification & the Road to Fingerprints

Industrial Revolution urbanisation increased prison populations & repeat-offender tracking challenges.
Pre-fingerprint solution: Non-standardised mugshots (1830s–1840s).
1888: Alphonse Bertillon creates modern standard mugshot + 11 anthropometric measurements (Bertillonage). Temporarily dominant.

Dr. Henry Faulds (1870s–1880s)

First to publish permanence & individuality evidence (Nature, 1880).
Introduced practical inking technique; proposed use in policing.
Syllabic classification: 21 consonants + 6 vowels → potential $\approx17\text{ trillion}$ categories (impractical; rejected by Scotland Yard 1886).

Sir Francis Galton (1892)

Book “Finger Prints” formalises permanence & uniqueness.
Tri-type notation (L, W, A) → sequence string (e.g., LAWLLWWLLW). Too coarse for large files but key conceptual precursor.

Juan Vucetich & “The New Argentine System”

1891: Advocates & implements fingerprint identification in Buenos Aires.
Expands Galton to 4 patterns: Arch (A/1), Internal Loop (-/2), External Loop (E/3), Whorl (น/4).
Primary = series (right) / section (left). Thumb = fundamental (right) or subclassification (left); remaining fingers = division/sub-division.
- Example given: Numerator $A1141$ (Right: A,A,A,W,A); Denominator $E2231$ (Left: E,2,2,E,1).
Publishes pamphlet (1896) & book “Dactiloscopia Comparada” (1904). Still used in many Spanish-speaking nations.

Henry Classification System (HCS)

Developed under Sir Edward Henry; mathematic underpinnings by Azizul Haque & Hem Chandra Bose (1897).
Purpose: Rapid filing & retrieval of known ten-print cards (not crime-scene latents).

Primary Classification Formula

$\text{Primary} = \frac{(\text{sum of even-finger whorl values})+1}{(\text{sum of odd-finger whorl values})+1}$

Finger order values (right thumb $\rightarrow$ left little finger): $16,8,4,2,1,16,8,4,2,1$ .
Any finger without a whorl contributes $0$ .
Yields $1024$ possible primary groupings.
- Example (all ten whorls): $\frac{32}{32}$ .
- Classroom example result: $\frac{14}{28}$ .

Secondary & Subsequent Extensions

Secondary: Patterns on #2 (Right Index) & #7 (Left Index) → capital letters (A,T,R,U,W). E.g., $\frac{14\;RW}{28}$ .
Rare non-index arches/tented/radial loops annotated lowercase after secondary; if in a thumb, placed before the primary (e.g., $a\;\frac{14\;RW}{28}$ ).
Sub-secondary: Loop ridge counts or whorl ridge tracing for remaining fingers appended right of secondary.

Performance & Limitations

Adopted India (1897), Scotland Yard (1900), widespread Europe.
Effective up to $\approx$ one million records; beyond that manual card sorting became unwieldy, spurring modifications & automated approaches.

Single-Fingerprint & Alternative Systems

1929 Battley Single-Fingerprint (Scotland Yard) aimed to match latent single impressions to known file; labour-intensive pre-digital.
Footprints:
- FBI system: Arch (O), Loop (L), Whorl (W); primary (capital) + secondary (lowercase); fraction $\frac{RF}{LF}$ .
- Chatterjee: Foot divided into 6 zones; alphabetic for Area-1 (ball) & numeric for Areas 2–6.
Palmprints (Western Australia, Liverpool, Denmark): All reference interdigital, thenar, hypothenar zones.

Automated Fingerprint Identification Systems (AFIS)

Definition: Computerised search of lawfully-obtained prints under Identification of Criminals Act (Canada) or equivalent.
Canadian implementation: CCRTIS (RCMP Ottawa); Livescan submission $\Rightarrow$ database search <5 min.
1960s global R&D (US, UK, Japan, France). Early resistance: cost, politics.

San Francisco Paradigm Shift (1983)

Mandated AFIS search for all identifiable crime-scene prints.
Formed 24/7 Crime Scene Investigations (CSI) unit.
Patrol officers must notify CSI on any felony with fingerprint potential.
CSIs trained to operate AFIS & search own cases.
Statistics gathered on AFIS hits to justify resources.

End-to-End AFIS Workflow (simplified)

Latent lifted → technician triages for quality/quantity.
Image digitised; minutiae, core, axes marked.
Algorithm compares against known database.
System returns ranked candidate list.
Operator/submitter manually compares (ACE-V), confirms or rejects.
Non-hit latents stored in crime-scene database for future searches.
Search scope: local $\rightarrow$ provincial $\rightarrow$ national $\rightarrow$ FBI/INTERPOL (time vs coverage trade-off).
Technician issues formal report regardless of hit/no-hit; manual suspect comparisons may still follow.

Key Acronyms

CPIC: Canadian Police Information Centre.
CNI: Criminal Name Index.
FPS#: Fingerprint Service Number.
CR: Criminal Record (charges & dispositions).

Technology Evolution & Vendors

Vendors: Cogent, Print Trak, Safran, Lockheed-Martin, etc.
Naming reflects generations: AFIS → IAFIS → RTAFIS → NGI.
Enhancements: faster algorithms, larger storage, improved UI, remote submission, higher resolution sensors.

AFIS in Canada

Initially physical courier to Ottawa; progressed to remote electronic terminals & real-time responses (Livescan).
Faster turnaround empowers frontline investigation (search warrants, arrests, bail decisions).

Next Generation Identification (NGI) – FBI

Launched 2011; modular multi-biometric platform superseding IAFIS.
Repository > $100$ million identities; exempt from US Privacy Act (ethical/privacy debate).
AFIT algorithm accuracy $99.6\%$ vs IAFIS $92\%$ .

NGI Functional Modules

RISC (Repository for Individuals of Special Concern): Mobile ID in <10 s; boosts officer safety.
National Palm Print System (NPPS): Nationwide latent/palm searching (2013).
Rap Back: Continuous monitoring – notifies agencies when individuals of trust are arrested (e.g., teachers).
Interstate Photo System (IPS): Contains mug shots + scars/marks/tattoos.
Facial Recognition Search: Queries $\approx30$ million images; returns ranked investigative leads (2010 error ≈ $20\%$ → improving).
Deceased Persons Identification (DPI): Matches post-mortem prints.
NGI Iris Service: Iris images linked to ten-print records; contactless rapid ID.

Significance, Ethics & Real-World Implications

Frequency data (arches $5\%$ , loops $65\%$ , whorls $35\%$ ) underpins probabilistic weighting; rare patterns (arches, radial loops) accelerate manual searches & affect courtroom testimony on uniqueness.
Transition from anthropometry to friction-ridge science revolutionised criminal justice, enabling objective repeat-offender tracking.
AFIS/NGI shift investigative timelines from weeks to minutes – transforms patrol, CSI, prosecution strategy.
Privacy concerns: NGI exemption, large-scale facial recognition, Rap Back continuous surveillance raise civil-liberty debates.
Technological accuracy $\ne$ infallibility: operator error, algorithm bias, poor-quality latents can produce false leads; ACE-V confirmation & transparency essential ethical safeguards.

Key Numerical & Formulaic References

Henry Primary groupings: $1024$ possible.
Whorl value assignment table: $[16,8,4,2,1,16,8,4,2,1]$ .
Example calculation (all whorls) $\frac{(16+8+4+2+1)+1}{(16+8+4+2+1)+1}=\frac{32}{32}$ .
Classroom exercise answer: $\frac{11}{28}$ (calculated via even/odd summations).
AFIS Livescan Canadian search time <5 min; NGI RISC <10 s.
AFIT fingerprint algorithm accuracy $99.6\%$ .

End-of-Lecture Checklist for Exam Preparation

Memorise pattern definitions & visual hallmarks (core, delta positions).
Practise Henry Primary computation steps & secondary/sub-secondary annotation rules.
Understand chronological progression: Bertillonage → Faulds → Galton → Vucetich → Henry → AFIS → NGI.
Be ready to discuss ethical/privacy implications of large biometric systems.
Know acronyms (CPIC, CCRTIS, CNI, FPS, RISC, NGI, ACE-V).
Review real-world examples (San Francisco AFIS rollout) for policy impact questions.