AXONAXON
Operator enrolling a person with a dedicated binocular iris scanner

Technology · Embedded capability

Iris recognition, verified in seconds.

A sub-five-second capture that scales to national populations — with the honest trade-off stated up front: the person has to be there, at a dedicated sensor.

Talk to the Identity TeamBinocular vs Monocular

Under 5 second capture

UIDAI device spec vs ~50s for ten-print

Scales past millions

NIST IREX III: FNIR below 2.5% at 3.9M

The short answer

Iris is the fastest biometric to capture and one of the hardest to fake at a distance — but it only works with the person standing in front of the sensor.

The iris is imaged under active near-infrared illumination, because melanin absorbs visible light and darker irides — the majority across many of the populations Axon serves — show little usable texture outside the infrared. That is also why a standard camera phone cannot do it: consumer cameras filter infrared light out.

Because it needs a dedicated sensor and a present subject, iris is not a remote-onboarding tool — pair it with 2D face matching wherever the person isn't physically there. But wherever presence is already guaranteed — enrolment desks, service centres, border posts — it is one of the fastest and most scalable biometrics available.

National ID enrolmentCivil registry dedupBorder & watchlist screeningField verificationVoter registration

Dedicated scanner

Purpose-built binocular capture for first enrolment.

The iris path starts with dedicated near-infrared hardware: controlled distance, shielded eyes and a repeatable presentation, instead of asking a phone camera to do work it was never designed for.

Both eyes presented to one scanner
Near-infrared capture in a controlled position
Field or desk workflow depending on the programme
Dedicated binocular iris scanner product image
COMET-M1 rugged handheld biometric device

Field-portable capture

One handheld, carried to the person.

COMET-M1 packs an optional iris capture module into the same rugged, IP67 handheld used for FAP30 fingerprint and 1:1/1:N face matching — a single ~500 g device that travels to the person instead of asking them to come to a fixed scanner.

IP67-rated, ~500 g single handheld unit
Captures iris alongside FAP30 fingerprint and face matching
Verifies against an existing enrolment out in the field, no desk required
View the COMET-M1

One eye, or two

Binocular for foundational enrolment. Monocular for the field.

The two configurations aren't interchangeable — they trade capture hardware for accuracy in opposite directions, so the right choice depends on whether this is a first enrolment or a repeat verification.

Both eyes, one presentation

Binocular capture

Captures both irides in a single presentation. Because the camera shields both eyes from ambient light, pupil dilation — the variable NIST identifies as a leading cause of iris matching error — stays in a narrower, more consistent range, and matching against two irides rather than one measurably improves accuracy.

Captures

Both eyes, one presentation

Accuracy gain

~2–3× lower false-non-match vs one eye

Requires

Dual-eye capture hardware

Recommended for

Foundational enrolment

Best fit

  • Foundational national ID and civil registry enrolment
  • Voter roll and large civil registration programmes
  • One-time enrolment events for large customer groups
  • Any programme where enrolment quality outweighs hardware cost

NIST tested the intuitive claim that two eyes double accuracy, and it does not hold: because image-quality issues tend to affect both eyes together, the real gain is a factor of two to three, not a square — and fusing both eyes typically raises the false-positive rate too. Binocular is still the right default for foundational enrolment; it is a measured accuracy gain in a single presentation, not a theoretical multiple.

One eye, any handheld

Monocular capture

Captures a single iris — the standard configuration for handheld and field devices, where size, weight and cost rule out a dedicated binocular camera. A monocular capture is compared 1:1 against an existing enrolment, or searched 1:N where only one eye is on file.

Captures

One eye per presentation

Accuracy

Strong single-eye baseline

Requires

Standard single-eye iris sensor

Recommended for

Field verification, handheld

Best fit

  • Field verification against an existing enrolment
  • Handheld devices where size and cost matter, such as COMET-M1
  • Repeat authentication once someone is already enrolled
  • Point-of-service and point-of-sale verification

A single eye still delivers a strong match on its own — NIST measured a best single-eye false-negative identification floor of about 1.5%, against 0.7% for two eyes. Monocular is the right choice once someone is already enrolled and is simply being verified, not for initial enrolment at foundational scale.

Verify one. Search millions.

Built for large populations and large customer groups.

1:1 and 1:N answer different questions, and a national programme needs both — verification at the point of service, deduplication at the point of enrolment.

1:1 · Verification

“Is this the same enrolled person?”

A single, real-time comparison against one existing iris enrolment — fast enough to run inline at a service counter, SIM swap or benefit disbursement.

Re-authentication at a service counter
SIM swap or account recovery
Confirming a field enrolment matches a prior record

1:N · Deduplication

“Has this person already been enrolled, under any identity?”

One iris searched against every record already on file — the control that stops someone enrolling twice under a different name. NIST's IREX III evaluation held the false-negative identification rate below 2.5% at a gallery of 3.9 million enrolled irides.

National ID and civil registry enrolment
Voter roll deduplication
Refugee and displaced-persons registration

Where the numbers stop

NIST's largest published iris gallery is 3.9 million enrolled irides, from 2012. Performance at tens or hundreds of millions — the scale of a national civil registry — is an extrapolation, and should be validated against the specific matching engine and population before it's relied on operationally.

National ID or civil registry enrolment

Foundational, one-time enrolment for large populations — the highest-assurance capture is worth the extra hardware.

1:N · Binocular

Voter roll registration

Deduplicate against the existing roll before a new registration is accepted, at national scale.

1:N · Binocular

Refugee and displaced-persons registration

Large customer groups processed quickly, often in difficult field conditions, where sub-five-second capture matters most.

1:N · Fast capture

Telecom customer registration at scale

High-volume retail registration where handheld, single-eye capture keeps throughput up at the point of sale.

1:N · Monocular

Benefits and welfare disbursement

Verify the recipient at the point of payment against their existing enrolment.

1:1

Border and watchlist screening

National programmes already run iris against watchlists at ports of entry.

1:N

Stated plainly

Where iris wins, and where it isn't the right fit.

A real trade-off, not a sales pitch: iris is fast, contactless and scales well precisely because it demands a controlled, in-person capture — the same property that rules it out for remote use.

Where iris is the right choice

  • Captures in under five seconds per presentation — India’s UIDAI sets this as the device specification for certified iris hardware — against a NIST-measured mean of 50 to 54 seconds for assisted ten-print fingerprint capture.
  • Fully contactless: no shared surface to clean between subjects, and no platen to wear down.
  • Not abraded by manual labour, and not subject to the age-related quality decline well documented in fingerprint capture.
  • Stable over the roughly ten-year horizons that have actually been measured — comfortably longer than most identity document validity periods.
  • Scales to large populations: NIST's IREX III evaluation held the false-negative identification rate below 2.5% at a gallery of 3.9 million enrolled irides.

Where it isn't the right fit

  • Requires the person to be physically present at a dedicated near-infrared sensor. Unlike 2D face matching, it cannot be enrolled from a selfie or an existing photograph.
  • Needs purpose-built capture hardware — a stock smartphone camera cannot do it, because consumer cameras filter out the infrared light iris capture depends on.
  • Still requires a cooperating subject at a known distance, not a passive, at-a-distance capture.
  • Presentation attack detection is effective against known attack types, but is not a solved problem against novel ones — pair it with liveness supervision or a second modality for high-value transactions.
  • NIST's largest published galleries stop at a few million enrolled irides; performance at tens of millions is an extrapolation to validate against the specific matcher.

Built in, not bolted on

One capture, matched into the record.

Axon integrates capture, quality gating, matching and business rules so the enrolment desk sees one coherent step and the registry receives one auditable record.

01

Capture

Binocular or monocular iris image under near-infrared illumination

02

Quality gate

ISO/IEC 29794-6 quality scoring before a comparison is attempted

03

Match

1:1 verification or 1:N search against the registry

04

Decide

Match, no-match, refer or step up

Operator workflow showing left and right iris images captured before continuing enrolment

Operator view

Capture quality is visible before the record moves on.

The workflow keeps left and right iris images visible at the enrolment step, with a clear recapture path before the operator commits the biometric record.

Left and right iris review in one step
Recapture before submit, not after failure
Clean handoff into summary, match and decision

COMET-M1

Captures iris in an easily portable, rugged handheld form — alongside FAP30 fingerprint and 1:1/1:N face matching — so field teams capture and verify iris without a fixed enrolment desk.

IRIS-1

Dedicated binocular iris scanner for controlled, dual-eye enrolment at desks, service centres and civil registry programmes.

Civil ABIS

Face, fingerprint and iris deduplicated together with score-level fusion across the full registry.

Start with the population

Let us specify the right capture hardware for your programme.

We'll work from population size, field conditions and connectivity — then specify binocular or monocular capture and the matching policy to match.

Talk to the Identity Team

Common questions

Iris recognition, without the fog.

Clear answers for product, fraud, security and operations teams.

Is iris recognition the same as retina scanning?+

No — they are different structures, imaged with different optics. The iris is the coloured ring at the front of the eye, captured from a short distance with a camera-style sensor. The retina is at the back of the eye and needs a close-range interior scan — the technology retina scanning has a reputation for being invasive that iris recognition does not share, because the two are frequently and incorrectly conflated.

Why can't a phone camera capture an iris? Doesn't every biometric work remotely now?+

Iris capture needs active near-infrared illumination, conventionally in the 700–900 nanometre band, because melanin absorbs visible light and darker irides — the majority in many of the populations Axon serves — show little usable texture outside the infrared. Consumer phone cameras carry infrared-blocking filters, so a stock smartphone cannot enrol an iris, and an ordinary visible-light photograph is not a usable enrolment source. That is why iris, unlike 2D face matching, is not a remote-onboarding tool — it requires the person at a dedicated sensor.

Does capturing both eyes double the accuracy?+

It's a common claim, and NIST tested it directly: it does not hold. Because image-quality issues — off-axis gaze, pupil dilation, cataracts — tend to affect both eyes together rather than independently, capturing two irides reduces the false-negative identification rate by a factor of roughly two to three, not by squaring it, and fusing both eyes typically increases the false-positive rate as well. Binocular capture is still the right default for foundational enrolment: it is a real, measured accuracy gain, delivered in a single presentation.

Is iris recognition spoof-proof?+

No biometric is. The credible attack vectors for iris are printed iris images, patterned contact lenses and prosthetic eyes; the countermeasure is presentation attack detection, tested under the ISO/IEC 30107 series. Independent testing shows PAD is effective against known, expected attack types, but the international LivDet-Iris competition has found that even leading algorithms accept a majority of presentations from a broad, partly unfamiliar set of attack instruments. For high-consequence transactions, we specify PAD tested to ISO/IEC 30107-3 and recommend supervision or a second modality rather than relying on iris PAD alone.

How stable is an iris over a lifetime?+

The iris forms in gestation and isn't subject to the wear that changes friction-ridge skin. NIST's longest operational study, using about a decade of border-crossing data, found no accuracy loss attributable to iris ageing over that period. Longer-term claims aren't something recognition data can support yet — no public dataset tracks the same irides across a full lifetime — but over the roughly ten-year horizons that have actually been measured, iris templates hold up well, comfortably longer than most identity documents remain valid.

What standards apply to iris data?+

ISO/IEC 19794-6:2011 specifies the iris image interchange format, and its successor, ISO/IEC 39794-6:2021, is what ICAO Doc 9303 now requires: inspection systems must be able to read it from 1 January 2026, and issuing states must use it from 1 January 2030. ISO/IEC 29794-6 covers how iris image quality is measured. In an ePassport, face is the required biometric and iris is optional — where a state includes it, ICAO requires storing the iris image itself, not just a matching template, specifically so the choice of matching vendor can change without re-enrolling the population.

Can the COMET-M1 capture and verify iris?+

Yes — the COMET-M1 captures iris in an easily portable, rugged handheld form factor, as an optional configuration alongside FAP30 fingerprint and 1:1/1:N face matching, so a single ~500 g device can both capture and verify iris in the field. For foundational, one-time national enrolment we recommend specifying a dedicated binocular scanner for the initial capture, then using COMET-M1 for field verification, re-authentication and mobile re-enrolment against that record.