Home / Products / Recognition Engines / Fingerprint Match Engine

ENGINE

Fingerprint Match Engine

Add fast, standards-based fingerprint matching to any application. The Cams Fingerprint Match Engine handles enrollment, 1:1 verification, and 1:N identification over ISO/IEC 19794-2 minutiae templates, with a two-stage search that pre-filters the gallery in SQL before scoring the survivors. A contactless fingerphoto path lets you match from an ordinary fingertip photo.

Get started Developer docs

Overview

ISO/IEC 19794-2 fingerprint enroll, 1:1 verify, and staged 1:N identify over a simple REST API.

A standalone fingerprint-matching microservice exposed at /api/biometric/v1/fp/. It stores ISO/IEC 19794-2:2005 single-finger-view minutiae templates per tenant and user, base64-encoded. Identification runs a staged 1:N pipeline: a coarse SQL pre-filter narrows candidates by finger position, quality, and pattern class, then a self-contained Bozorth-style minutiae matcher scores the survivors against a configurable threshold (default 72, about 12 matched minutiae — the conventional forensic minimum). The matcher sits behind a pluggable seam so an ISO comparison, NBIS/Bozorth3, or a commercial SDK can be wired in without touching the service.

ENGINERESTHTTPSJSONC++

POST /api/biometric/v1/fp/verify (1:1)

curl .../api/biometric/v1/fp/verify \
  -H "sn: $SN" -H "auth-token: $TOKEN" \
  -d '{ "user_id":"1001",
        "iso_template":"<base64 ISO 19794-2>" }'
{ "ok":true, "match":true, "score":81, "user_id":"1001" }

Capabilities

What it does

Enroll a Finger

Store an ISO 19794-2 minutiae template per user with finger position, quality, and pattern class for 1:N pre-filtering.

Verify (1:1)

Match a probe template against a specific user's enrolled fingers and return a 0-100 score and match flag.

Identify (1:N)

Staged search across the tenant gallery: SQL pre-filter then minutiae fine-match, returning the best candidate above threshold.

Standards-Based Templates

Works on the open ISO/IEC 19794-2:2005 standard; non-conforming templates are rejected, not silently scored as garbage.

Contactless Fingerphoto

Accepts a fingertip photo, converts it to an ISO template, and enrolls or identifies it through the same matcher.

Pluggable Matcher

Swap the scoring engine (ISO, NBIS/Bozorth3, or commercial) behind a single seam.

Features

Everything included

Three REST routes — enroll, verify (1:1), and identify (1:N) — at /api/biometric/v1/fp/
ISO/IEC 19794-2:2005 single-finger-view minutiae templates, base64-encoded and multi-tenant
Staged 1:N: coarse SQL pre-filter on finger position, quality, and pattern class, then fine minutiae match
Self-contained Bozorth-style minutiae matcher with translation-invariant pairwise geometry, no external NIST dependency
Configurable match threshold (default 72, about 12 matched minutiae)
Pluggable matcher seam to drop in ISO comparison, NBIS/Bozorth3, or a commercial matcher with no other changes
Multi-threaded candidate matching at ~0.09 ms per comparison (about 10.7k comparisons/sec per core)
Contactless fingerphoto enroll and identify path: a fingertip photo is converted to an ISO template and matched through the same engine
Clear in-band failure reasons (template not parseable, no candidates) rather than silent mis-scoring

Build with Fingerprint Match Engine

Grab a key, read the docs, and ship. Our team helps with your first integration.

Talk to an engineer WhatsApp

FAQ

Common questions

What fingerprint template format does it use?

ISO/IEC 19794-2:2005 single-finger-view minutiae records, base64-encoded. ANSI-378 and the 2011 ISO variant are not parsed and are rejected with a clear reason rather than silently mis-scored.

How does 1:N fingerprint identification work?

In two stages: a SQL coarse pre-filter narrows candidates by finger position, quality, and pattern class, then a Bozorth-style minutiae matcher scores the survivors and returns the best match above the configurable threshold.

Can it detect fake fingers?

Not at the server. The engine only sees the minutiae template, so liveness must come from the capture sensor or client SDK. Cams recommends LFD-certified scanners for fake-finger resistance.

Can it match a fingerprint from a phone photo?

Yes. A contactless fingerphoto path converts a fingertip photo into an ISO 19794-2 template the engine then enrolls or identifies.