ShieldPST.ai · Technology Explainer Series

Facial Recognition Technology

How facial recognition converts facial imagery into mathematical representations, compares a probe image against one or many reference images, produces candidate results, and assists law-enforcement investigations—and what agencies should understand about image quality, thresholds, human review, demographic effects, misidentification, watchlists, evidence, privacy, disclosure, and governance.

1:N Search Probe Face Against a Gallery
Output Candidate · Not Positive Identification
Core Safeguard Independent Human Verification

What this explainer does

Facial recognition technology uses computational analysis of facial imagery to compare faces. Law-enforcement systems may compare a photograph of an unknown person against a gallery of known photographs and return possible candidates for further investigation.

The technology can help investigators develop leads when a person's face appears in surveillance video, a photograph, digital evidence, or other imagery but the person's identity is unknown.

Facial recognition can be powerful, but its output is probabilistic. A returned candidate should not be treated as automatic proof of identity. Image quality, gallery composition, algorithm performance, thresholds, demographic effects, and human review can all affect the process.

The most important rule

A facial-recognition candidate is an investigative lead—not a positive identification.

Identity should be confirmed through independent evidence before arrest, charging, warrant affidavits, or other consequential action.

1. Overview

Facial recognition technology uses algorithms to measure and compare facial characteristics in digital images.

In a typical investigative application, law enforcement possesses an image of an unknown person. That image is submitted to a facial-recognition system and compared against a gallery containing photographs associated with known or previously identified individuals.

The system calculates similarity and may return one or more candidate photographs. A trained reviewer then evaluates those candidates and the underlying imagery before investigators determine whether further investigation is warranted.

Central Concept The process should be understood as: image → algorithmic comparison → candidate result → human evaluation → independent investigation → identity confirmation . Skipping the later steps creates substantial risk.

2. Face Detection, Verification, and Identification Are Different

Function Question Example
Face Detection Is there a face in this image? Software finds faces so they can be blurred, cropped, or analyzed.
1:1 Verification Are these two images likely the same person? Comparing a live image with the photograph associated with a claimed identity.
1:N Identification Who might this unknown face be? Comparing a surveillance image against thousands or millions of gallery images.
Terminology Rule Face detection is not facial identification. A system capable of finding faces in video is not necessarily identifying who those people are.

3. How a 1:N Facial Recognition Search Works

1. Obtain Image Investigators select a usable image of the unknown person
2. Detect Face Software locates and processes facial characteristics
3. Create Template Facial information is converted into a mathematical representation
4. Search Gallery Probe template is compared against enrolled gallery images
5. Rank Candidates Possible matches are ordered by algorithmic similarity
6. Investigate Human review and independent evidence test the candidate identity

4. The Probe Image

The image submitted for comparison is commonly called the probe image.

Probe quality can dramatically affect performance. A clear front-facing booking photograph presents a very different technical problem from a blurry surveillance frame showing part of a face at an angle.

Resolution

Too few facial pixels can reduce usable identifying information.

Pose

Profile views or extreme head angles can reduce comparability.

Lighting

Strong shadows, backlighting, darkness, or overexposure can alter facial detail.

Occlusion

Masks, glasses, hats, hair, hands, or other objects can obscure features.

Motion Blur

Fast movement or poor camera exposure can degrade facial structure.

Compression

Repeated screenshots, messaging applications, and video compression can eliminate detail.

6. Similarity Scores Are Not Probabilities of Guilt

Face-recognition systems typically calculate a numerical measure of similarity between facial representations.

The meaning of the resulting score depends on the algorithm, threshold, gallery, configuration, and vendor.

Critical Caution A score such as “92” or “0.87” should not automatically be described as “92% certain this is the person.” Similarity score, algorithm confidence, match threshold, and statistical probability are different concepts.

7. Candidate Lists

Many law-enforcement 1:N systems return a ranked candidate list rather than one definitive identity.

The algorithm is effectively saying: these gallery images are among those most similar to the probe under this system's comparison method.

Rank

A candidate may appear first, second, fifth, or elsewhere in the returned list.

Threshold

The system may exclude candidates falling below a configured similarity threshold.

Gallery Size

Searching millions of images creates a different false-candidate environment than comparing two known images.

Operational Rule A candidate list is a lead-generation tool. Investigators should independently determine whether the candidate actually fits the case before treating identity as established.

8. Image Quality Can Dominate the Result

Agencies sometimes focus heavily on algorithm accuracy while underestimating the importance of the probe image itself.

Surveillance imagery can be particularly difficult because cameras are frequently mounted for scene coverage rather than biometric identification.

Camera Height

Overhead cameras may capture the top of the head more clearly than facial features.

Distance

A face occupying only a small portion of the frame contains limited information.

Frame Selection

The best frame may occur before or after the moment investigators initially notice.

Quality Control Agencies should establish standards governing when an image is too poor for a facial-recognition search rather than assuming every visible face should be submitted.

9. What Does “Accuracy” Mean?

Facial-recognition performance cannot responsibly be reduced to one universal accuracy percentage.

Measure Question
False Positive How often does the system incorrectly associate different people?
False Negative How often does the system fail to recognize the same person?
False Positive Identification Rate In 1:N searching, how often does a non-matching search yield an incorrect candidate above threshold?
False Negative Identification Rate How often does the true person fail to appear appropriately in the candidate result?
Rank Performance How often does the correct individual appear within the first 1, 5, 10, or other candidate positions?
Operational Error How often do algorithmic, human, procedural, or investigative errors ultimately produce a wrong identification?
Metrics Principle Ask: Which algorithm? Which version? Which dataset? 1:1 or 1:N? What threshold? What gallery size? What image quality? What demographic groups? What operational workflow? Without that context, an “accuracy rate” has limited meaning.

10. Demographic Effects

Independent testing has demonstrated that facial-recognition algorithms can exhibit different error rates across demographic groups. Performance varies significantly among algorithms.

These differences are not uniform across every vendor or every type of error. Agencies therefore should evaluate the particular algorithm, version, operating environment, and use case they actually deploy.

Race / Population Group

Error differentials have been documented across population groups in facial-recognition evaluations.

Sex

Performance differences can occur across sex categories depending on algorithm and dataset.

Age

Age and aging can affect comparison performance, particularly when reference and probe images are separated in time.

Bias Principle Do not ask only: “Does facial recognition have demographic bias?” Ask: “What are the measured error rates of this algorithm, for this task, under these conditions, across relevant populations?”

11. Human Review Is Essential—but Humans Can Also Make Errors

Human review is an essential safeguard because algorithms identify similarity rather than proving identity.

But human review must be structured. A reviewer can be influenced by expectations, rank order, contextual information, confirmation bias, or visual similarity among different people.

Trained Examiner

Reviewers should understand facial comparison and system limitations.

Quality Assessment

Reviewers should evaluate whether the probe is actually suitable for comparison.

Documented Conclusion

The review process and basis for advancing or rejecting a candidate should be documented.

Human-in-the-Loop Principle Human review should be meaningful review—not ceremonial approval of the algorithm's top-ranked candidate.

12. How Misidentification Can Happen

Poor Probe Low-quality image contains limited facial information
Candidate Returned Algorithm identifies a visually similar gallery image
Reviewer Anchors Human reviewer accepts the candidate too readily
Investigator Anchors Subsequent evidence is interpreted through candidate identity
Weak Corroboration Ambiguous facts are treated as confirmation
Wrong Action Stop, arrest, search, or charging follows misidentification
Misidentification Risk The greatest danger may not be the algorithm making one incorrect comparison. It may be automation bias causing every later investigator to treat that tentative candidate as the established suspect.

13. Watchlists and Real-Time Recognition

Some systems can compare faces observed by cameras against a predefined watchlist rather than waiting for investigators to submit a photograph after an event.

This changes the risk profile substantially.

Continuous Scanning

People may be compared automatically without being individually suspected of wrongdoing.

Watchlist Quality

Outdated, erroneous, overbroad, or poorly sourced entries can generate inappropriate alerts.

Immediate Action

Real-time alerts can create pressure to act before meaningful verification occurs.

14. Facial Recognition in an RTCC

Facial recognition becomes particularly consequential when integrated into a Real-Time Crime Center.

An analyst could potentially receive imagery from a camera, perform facial recognition, obtain a candidate, search agency records, examine associated vehicles, check ALPR history, and communicate a possible identity to officers within minutes.

RTCC Safeguard RTCC speed increases the need for a defined verification workflow. Field officers should be told whether an identity is: confirmed, strongly corroborated, facial-recognition candidate only, or otherwise uncertain.

15. Facial Recognition and Body-Worn Cameras

Body-worn cameras create an enormous volume of facial imagery. Technically, that creates opportunities for both retrospective and potentially real-time facial analysis.

But adding facial recognition to BWC systems can transform a recording device into an identification sensor capable of analyzing large numbers of police-public encounters.

Function-Creep Question Agencies should separately decide whether facial recognition may be applied to BWC footage rather than assuming that permission to record an encounter automatically authorizes biometric identification of everyone appearing in the recording.

16. Fourth Amendment Considerations

There is no single Supreme Court decision establishing one comprehensive Fourth Amendment rule for all law-enforcement facial-recognition uses.

The constitutional analysis can depend on the source of the probe image, how the gallery was created, whether other protected information is accessed, the surveillance context, duration, location, and what officers do after receiving a candidate result.

Public Surveillance Image

Raises different questions from imagery obtained through a separate unlawful search.

Government Database

Authority to maintain or search particular photographs can depend on statutes, policy, and source.

Resulting Stop or Arrest

The facial-recognition lead does not eliminate the applicable requirement for reasonable suspicion or probable cause.

17. First Amendment and Protected Activity

Facial recognition can implicate expressive and associational interests when used at demonstrations, political events, houses of worship, journalistic activity, organizational meetings, or similar contexts.

Protected-Activity Rule Facial recognition should not be used solely because a person is exercising protected First Amendment rights. Policies should establish heightened approval and purpose requirements for biometric identification associated with protests, political gatherings, religion, journalism, or other protected activity.

18. The Candidate Is Not the Evidence That Proves Identity

Once facial recognition develops a candidate, investigators should seek independent evidence.

Physical Characteristics

Height, build, tattoos, scars, clothing, or other features may corroborate or exclude the candidate.

Location Evidence

Phone, ALPR, video, witnesses, transactions, or other information may connect the individual to the relevant place.

Investigative Records

Known relationships, vehicles, addresses, conduct, admissions, or other evidence may independently support identity.

Report-Writing Rule Instead of writing: “Facial recognition identified Smith as the suspect.” use precise language describing that the search returned Smith as a candidate or investigative lead and then separately describe the evidence that independently established or corroborated identity.

19. Discovery, Disclosure, and Litigation

Facial recognition can create discoverable or potentially relevant records beyond the final candidate name.

Record Why It May Matter
Original Probe Shows the imagery actually available for comparison.
Processed Probe Documents cropping, enhancement, rotation, or other preprocessing.
Candidate List Shows which identities the system returned and their rank.
Scores May document algorithmic similarity information.
Gallery Identifies what population was searched.
Algorithm / Version Establishes which system produced the result.
Reviewer Notes Shows human evaluation of returned candidates.
Audit Log Documents search, user, date, system, and related activity.

20. How Agencies Should Validate Facial Recognition

Independent laboratory testing is valuable, but agencies also need to determine whether the technology is appropriate for their intended operational use.

1. Define Use Investigative leads, watchlists, verification, RTCC, or another purpose
2. Review Testing Examine independent evaluation of the actual algorithm
3. Test Images Use realistic surveillance and agency image conditions
4. Measure Errors Evaluate false positives, false negatives, rank, and demographics
5. Validate Workflow Test analysts, reviewers, reports, and investigative confirmation
6. Revalidate Repeat after major algorithm, gallery, threshold, or policy changes
Validation Principle Evaluate the system + image + gallery + threshold + human reviewer + investigative workflow. Real-world failure can occur at any of those stages.

21. Procurement Questions

Issue Agency Question
Algorithm Which exact algorithm and version will the agency use?
Independent Testing Has the algorithm been evaluated by NIST or another credible independent body?
Gallery Who supplies the reference images and what populations are included?
Threshold Who determines the similarity threshold and can it be changed?
Model Changes Can the vendor change the algorithm without notifying the agency?
Search Records Are probe images, candidate lists, scores, and audit records preserved?
Training Use May vendor systems use agency images to train or improve models?
Subprocessors Which third parties can access biometric data?
Data Ownership Who owns templates, probe images, candidate information, and logs?
Audit Rights Can the agency independently audit search use and system performance?

22. Agency Governance Framework

Authorized Purposes

Specify when facial recognition may and may not be used.

Seriousness Threshold

Consider whether use should be limited by offense, risk, investigative need, or other defined criteria.

Probe Standards

Require minimum image-quality assessment before search.

Trained Reviewers

Restrict candidate evaluation to appropriately trained personnel.

Lead-Only Rule

Expressly state that a candidate result is not positive identification.

Independent Corroboration

Require additional evidence before consequential enforcement action.

Watchlist Controls

Establish entry criteria, expiration, review, correction, and deletion procedures.

Protected Activity

Apply heightened safeguards to protests, religion, politics, journalism, and other First Amendment activity.

Gallery Governance

Document legal authority, source, retention, and permitted uses of searchable photographs.

Audit Logs

Record who searched what image, when, why, and what result was produced.

Discovery

Coordinate preservation and disclosure requirements with prosecutors.

Periodic Revalidation

Reassess performance after material changes to algorithms, galleries, thresholds, vendors, or workflows.

23. Questions Every Agency Should Answer

What exact facial-recognition systems does the agency use?
Does the agency own the system or use another agency's or vendor's service?
What algorithm and version performs the comparison?
Has that algorithm been independently evaluated?
Is the use 1:1 verification, 1:N identification, or both?
Does the agency use face detection without identification?
What photographs may be submitted as probes?
What minimum probe-image quality is required?
May analysts enhance or alter probe images?
Are original and processed versions preserved?
What gallery or galleries are searched?
Where do gallery photographs originate?
Do galleries contain only criminal booking images?
Are driver's-license or other civil photographs searchable?
What legal authority governs gallery use?
What similarity threshold is used?
Who may change that threshold?
How many candidates are returned?
Are candidate scores preserved?
Who evaluates candidate lists?
What facial-comparison training is required?
Is a second reviewer required in high-risk cases?
Does policy expressly state that a candidate is not positive identification?
What independent evidence is required before arrest?
Can a facial-recognition candidate alone support a warrant affidavit?
How must the use of facial recognition be described in police reports?
Are prosecutors notified when facial recognition materially contributes to a case?
What FRT records are retained for discovery?
Can the agency reconstruct a historical search?
Are every user's searches logged?
Are random audits performed?
Does the agency measure false candidate rates?
Does the agency monitor demographic performance?
Does the agency use real-time facial recognition?
Does the agency maintain facial-recognition watchlists?
Who may be placed on a watchlist?
How long may a person remain on a watchlist?
How are incorrect or outdated watchlist records corrected?
May facial recognition be used during demonstrations?
What heightened approval applies to First Amendment activity?
Can facial recognition be applied to body-worn camera footage?
Can RTCC personnel conduct facial-recognition searches?
Can candidate results be sent directly to field officers?
How must uncertainty be communicated to field officers?
Can private-camera feeds be subjected to facial recognition?
May vendor systems use agency images for model training?
Where are biometric templates processed and stored?
Which subcontractors or subprocessors have access?
Does the vendor notify the agency before algorithm changes?
What event triggers system revalidation?

24. Where Facial Recognition Is Going

Real-Time Identification

Live camera streams may increasingly be compared against authorized watchlists.

RTCC Integration

Candidate identities may be rapidly correlated with vehicles, records, locations, and other sensors.

Video Search

Systems may search hours or days of archived video for appearances of a particular face.

Cross-Camera Tracking

Facial features may be combined with clothing, gait, vehicles, and other attributes to reconstruct movement.

Body-Worn Camera Analytics

Facial identification could potentially be layered onto police-public encounter footage.

Multimodal Identification

AI may increasingly combine faces with voice, gait, vehicles, location, and other attributes.

Future-Looking Principle The largest governance change may come when facial recognition shifts from: “Who is the unknown person in this crime-scene image?” to: “Tell us whenever this person appears anywhere in the camera network.” Those are fundamentally different surveillance capabilities and should not be governed as though they were identical.

25. Key Terms

Facial Recognition Technology (FRT) Technology used to compare facial imagery for verification or identification purposes.
Face Detection Determining whether an image contains a face without identifying the person.
1:1 Verification Comparing two facial images to determine whether they are likely from the same person.
1:N Identification Comparing one probe image against a gallery containing many reference images.
Probe Image The image of the unknown or questioned person submitted for comparison.
Gallery The collection of reference facial images against which the probe is searched.
Face Template Mathematical representation of facial characteristics used for algorithmic comparison.
Candidate A person returned by a facial-recognition search as potentially similar to the probe.
Candidate List Ranked group of possible matches returned by a 1:N facial-recognition search.
Similarity Score System-specific measure describing algorithmic similarity between facial representations.
Threshold Configured score or decision boundary used to determine whether a comparison result is returned or accepted.
False Positive Incorrectly treating images of different people as sufficiently similar.
False Negative Failure to recognize two images of the same person.
FPIR False Positive Identification Rate used in evaluation of one-to-many identification systems.
FNIR False Negative Identification Rate measuring failures to return the correct identity appropriately in a 1:N search.
Watchlist Predefined collection of identities against which incoming facial imagery may be compared.
Automation Bias Tendency to place excessive trust in a computer-generated result or recommendation.
Demographic Differential Difference in measured algorithm performance across demographic groups.
Human-in-the-Loop Workflow requiring meaningful human review before consequential reliance on automated output.
Biometric Information Data derived from physical or behavioral characteristics used for recognition or identification.

26. Related ShieldPST.ai Resources

Real-Time Crime Centers

Integrated cameras, ALPR, databases, sensors, AI, analysts, and operational intelligence.

Open explainer →
Body-Worn Camera Analytics

AI-assisted video analysis, search, transcription, redaction, report generation, and evidentiary integrity.

Open explainer →
Social Media & OSINT

Public-source investigation, online identity, automated analysis, and investigative verification.

Open explainer →
Digital Evidence Center

Preservation, metadata, discovery, authentication, and digital evidence governance.

Open resource →
Police Technology Case Law Center

Research Fourth Amendment and emerging police-technology decisions.

Browse case library →
Technology Explainers

Return to the Shield Technology Reference Library.

Browse explainers →

27. Selected Primary and Authoritative Sources

National Institute of Standards and Technology — Face Recognition Technology Evaluation (FRTE), 1:N Identification
NIST's ongoing independent evaluation of one-to-many facial-recognition algorithms, including false-positive and false-negative identification performance.
Review NIST 1:N evaluation
National Institute of Standards and Technology — Face Recognition Technology Evaluation, 1:1 Verification
NIST's evaluation program for one-to-one facial verification algorithms.
Review NIST 1:1 evaluation
NIST — Demographic Effects in Face Recognition
NIST research and ongoing evaluation concerning measured performance differences across age, sex, race, and other demographic categories.
Review demographic evaluations
NISTIR 8280 — Face Recognition Vendor Test Part 3: Demographic Effects
Major NIST study evaluating demographic differentials across facial-recognition algorithms and image collections.
Review NIST report
FBI — Next Generation Identification: Interstate Photo System Facial Recognition Search
FBI description of the NGI facial-recognition search service, which provides authorized law enforcement with ranked potential candidates as investigative leads.
Review FBI NGI information
U.S. Department of Justice — Artificial Intelligence and Criminal Justice
DOJ report discussing law-enforcement facial-recognition use and safeguards, including the principle that facial-recognition results alone may not serve as sole proof of identity and that identity must be confirmed through other investigation or analysis.
Review DOJ report
U.S. Government Accountability Office — Facial Recognition Services: Federal Law Enforcement Agencies Should Take Actions to Implement Training and Policies for Civil Liberties
GAO review of federal law-enforcement use of facial-recognition services, policy, training, privacy, civil-rights, and civil-liberties safeguards.
Review GAO report
U.S. Government Accountability Office — Facial Recognition Services: Federal Law Enforcement Agencies Should Take Actions to Implement Training Requirements
GAO review addressing federal facial-recognition use and the FBI's Facial Recognition Technology Use Policy Directive and training requirements.
Review GAO report
U.S. Government Accountability Office — Facial Recognition Technology: Federal Law Enforcement Agencies Should Better Assess Privacy and Other Risks
GAO examination of agency-owned and external facial-recognition systems, tracking, privacy, accuracy, and risk assessment.
Review GAO report

28. Key Takeaways

Bottom Line
  1. Face detection, one-to-one verification, and one-to-many identification are different technologies and should not be treated as interchangeable.
  2. Law-enforcement facial identification commonly involves comparing a probe photograph of an unknown person against a gallery of known photographs.
  3. A facial-recognition system normally measures similarity; it does not independently prove identity.
  4. A returned candidate should be treated as an investigative lead rather than a positive identification.
  5. Investigators should independently corroborate identity before arrest, charging, search warrants, or other consequential action.
  6. Probe-image quality can materially affect performance. Resolution, pose, lighting, motion, occlusion, compression, and camera angle all matter.
  7. The gallery matters as much as the algorithm. Agencies should know whose photographs are searchable and under what authority.
  8. Similarity scores should not be translated casually into percentages of certainty that a person is the suspect.
  9. Accuracy should be evaluated separately for false positives, false negatives, one-to-one verification, one-to-many identification, gallery size, rank, and operational conditions.
  10. Independent testing has documented demographic differentials in facial-recognition performance, and those differences vary substantially among algorithms.
  11. Human review is essential but can itself introduce confirmation bias or automation bias unless reviewers are trained and follow structured procedures.
  12. Real-time watchlist recognition creates greater privacy and operational risk than a single post-event investigative search and should receive separate governance treatment.
  13. RTCC integration can compress the path from camera image to facial-recognition candidate to field action into minutes, increasing the importance of verification and clear communication of uncertainty.
  14. Agencies should separately decide whether facial recognition may be applied to body-worn camera recordings.
  15. Facial-recognition use does not eliminate the constitutional requirements governing stops, searches, arrests, home entries, or other enforcement actions.
  16. Policies should include heightened safeguards when facial recognition intersects with protests, religion, journalism, political activity, or other protected First Amendment conduct.
  17. Agencies and prosecutors should preserve enough information to reconstruct the facial-recognition process during discovery and litigation.
  18. The next major governance challenge is the shift from post-event identification of an unknown suspect to persistent, real-time identification across networked cameras.

ShieldPST.ai · Technology Explainer Series

This explainer is provided for training and general informational purposes. It is not legal advice and does not replace current review of controlling federal and state law, state constitutional provisions, biometric privacy statutes, First Amendment requirements, Fourth Amendment requirements, public-records law, discovery obligations, evidentiary rules, criminal-intelligence requirements, agency policy, local restrictions, vendor capabilities, algorithm evaluations, prosecutorial guidance, contractual requirements, or consultation with agency counsel. Facial-recognition technology, artificial intelligence, biometric law, and state and local regulation remain rapidly developing.

© 2026 Shield Public Safety Training. All rights reserved. · Reviewed August 10, 2026.