Facial Recognition Technology
What Is Facial Recognition Technology?
South Wales Police uses facial recognition technology to prevent and detect crime and help protect the vulnerable.
Facial Recognition is a technology capable of comparing a human face from a digital image against a database of faces.
Guide to all Facial Recognition Terminology and definitions
How does Facial Recognition Technology work?
It analyses key facial features and generates a mathematical representation of these features. It then compares them against the mathematical representation of known faces in a database, generating possible matches.
South Wales Police uses Facial Recognition Technology in the following cases:
- Live Facial Recognition (LFR) compares a live camera feed of faces against a predetermined watchlist to find a possible match that generates an alert.
- Retrospective Facial Recognition (RFR) is a post event use of facial recognition technology, which compares still images of faces of unknown subjects against a reference image database in order to identify them.
- Operator Initiated Facial Recognition (OIFR) is a mobile phone use of FRT technology, which compares a photograph of a person’s face taken on a mobile phone to the predetermined watchlist to assist an officer to identify a subject.
Whenever we use Facial Recognition Technology, we recognise there is an ongoing need to balance privacy and protection concerns, in addition to considering its legal and ethical implications.
Please contact us as [email protected] with any further enquiries.
Operational testing by South Wales Police
Why did we test?
The Metropolitan Police Service (MPS) and South Wales Police (SWP) tested Facial Recognition Technology (FRT) with the National Physical Laboratory (NPL). The NPL is a world-leading centre of excellence that provides cutting-edge measurement in science, engineering and technology. Thanks to previous testing by the National Institute of Standards and Technology (NIST) the MPS and SWP knew that their FRT used a high performing algorithm; the aim of the testing was to develop an in-depth understanding of the performance of the algorithms when it was being used in an operational environment. The three policing use cases were:
- Live Facial Recognition (LFR)
- Retrospective Facial Recognition (RFR)
- Operator Initiated Facial Recognition (OIFR)
The NPL test plan was specifically designed to help identify any impact this technology may have on any protected characteristics, in particular race, age and sex.
What do the test results tell us about the algorithm?
The NPL report gives us an impartial, scientifically underpinned and evidence-based analysis of the performance of the facial recognition algorithm currently used by the MPS and SWP. It tells us:
- We now better understand the demographic performance of our LFR system. There are settings our algorithm can be operated at where there is no statistical significance between demographic performance.
- that there was no demographic performance variation for RFR; and
- that there was no demographic performance variation for OIFR
The results from the tests assist the MPS and SWP with further understanding how to use FRT fairly in order to prevent and detect crime, safeguard national security and keep people safe.
The full results are presented in the NPL’s commissioned report: ‘Facial Recognition Technology in Law Enforcement Equitability Study’
Further information about the testing can be found in the Test Strategy for Facial Recognition Technology in Law Enforcement Equitability Study.
What is Live Facial Recognition (LFR)?
Live Facial Recognition (LFR) compares a live camera feed of faces against a predetermined watchlist in real-time. It is used to locate persons of interest by generating an alert when a possible match is found.
How is Facial Recognition Technology used?
South Wales Police currently utilises NEC’s NeoFace facial recognition technology.
Before using LFR, a watchlist of offenders is created, that are wanted by the police and courts. The watchlist also includes people who may pose a risk of harm to themselves or others.
LFR cameras target an area and the images are streamed to the live facial recognition technology. The images are compared against the images in the watchlist. When the technology finds a possible match an alert is generated.
An officer then compares the camera image to the person they see and decides whether to speak to the person.
We will always explain why we have chosen to speak with someone, and give them an informational leaflet with contact details if they have further questions. Persons who are not included on a watchlist cannot be identified.
We will delete the images with alerts immediately after using them or within 24 hours. The images and biometric data of those who don’t cause an alert are automatically and immediately deleted. We record the CCTV footage that is used by the LFR technology and we keep it for 31 days.
South Wales Police's Privacy Notice explains how we collect, store, use, disclose, retain and destroy personal data.
When is Live Facial Recognition (LFR) Technology Used?
Live Facial Recognition technology is considered an efficient and effective policing tactic to prevent and detect crime, and protect the most vulnerable in our society.
We typically use the technology at public events and crowded public spaces and will be deployed to aid policing operations where we have intelligence that supports its use.
How the police inform the public it’s using LFR
South Wales Police will do everything reasonably possible to inform the public where we are using Facial Recognition technology for example:
- Online before it is used
- South Wales Police website where the results are published of when LFR is used
- Information leaflets for the public
- Posters and signs to include QR codes to make people aware LFR is being used and where they can go to find further information
- Police officers are available to explain to the public how LFR works and why we are using it
FAQs on LFR
If you have any questions about the use of Live Facial Recognition technology, take a look at the frequently asked questions (FAQs).
What is Retrospective Facial Recognition (RFR)?
Retrospective Facial Recognition (RFR) technology compares still images of faces of unknown people against a reference image database in order to identify them after an event.
How Retrospective Facial Recognition works
South Wales Police currently utilises NEC’s NeoFace facial recognition technology.
RFR is utilised post event as part of a criminal investigation. Images are typically supplied from CCTV, mobile phone footage or social media. These images are then compared against our custody images.
The reference database that is utilised by South Wales Police are custody images from South Wales and Gwent Police. These contain in excess of 600,000 images.
After a search is made, the technology reorders the reference image database from the most likely to the least likely possible match.
An operator will typically review the top 200 possible matches to decide whether a match has been made. If the operator determines a match has been made, they will inform the investigating officer. The investigating officer will review the match and add the person to the investigation as a suspect.
When is Retrospective Facial Recognition used?
RFR technology is used to assist with the investigation of a criminal offence in order to efficiently and effectively identify a suspect. When a suspect of a crime remains unknown, RFR will be used.
Successful outcomes of crimes using RFR
View the news catalogue where Retrospective Facial Recognition Technology has been used to provide a positive match of a criminal leading to a sentence.
Retrospective Facial Recognition documents
Documents relevant to South Wales Police's use of Retrospective Facial Recognition documents.
What is the Operator Initiated Facial Recognition (OIFR)?
The OIFR App is a mobile phone App of Facial Recognition technology. It compares a photograph of a person’s face taken on a mobile phone with facial images contained in a predetermined watchlist(s) in order to assist an officer to identify a subject.
How does the Operator Initiated Facial Recognition (OIFR) App work?
An image will be captured on an officer’s mobile phone and compared against a watchlist(s) selected by the officer. The watchlist available to the officer includes custody images held by South Wales and Gwent Police and missing persons from the South Wales Police area.
After a search is made, the technology reorders the chosen watchlist(s) from the most to least likely possible match. The six most likely possible matches are returned to the officer’s mobile phone. The officer will review the top six possible matches and decide whether a match has been made. If a match has been made it will then be possible to carry out further checks for the subject against police systems.
We will always explain to the person when we use the OIFR App and inform them how they can contact us to ask further questions.
Persons who are not included on a watchlist cannot be identified. The image, captured on the mobile device, and biometric data are automatically and immediately deleted after a search is carried out.
When is the Facial Recognition (OIFR) App used?
The OIFR App will not be used to replace traditional means of identification, such as having a conversation with the individual who gives their name or provides identification documents which are checked against police systems to identify a subject.
Wherever possible, the OIFR App will only be used after an interaction has occurred between the officer and the subject.
The OIFR App does not replace the existing RFR process but instead is used as an ‘on-street’ intelligence tool to assist the officers in identifying an unknown person.
Use of the OIFR App will only occur when the identity of a subject is not known and when there are reasonable grounds for its use.
Reasons for using Facial Recognition (OIFR) App: -
- The subject is unable to provide their details because they are deceased, unconscious, have mental health or age barriers or incapable due to drink or drugs.
- The subject has refused to provide their details
- It is reasonably suspected that the subject has provided false details.
AND when one of the following grounds apply. When the subject is:-
- Deceased
- Reported as a missing person
- Suspected that they have committed an offence, wanted by the courts, subject to bail conditions or a court order.
- Presenting a risk of harm to themselves or others
Deployments
Find out about the deployments of live facial recognition technology.
Previous deployments
Download a pdf of the results of all deployments for Live Facial Recognition
What we need to capture post deployment:
- True Alerts
- Confirmed True Alerts
- False Alerts
- Confirmed False Alerts
- False Alert Rate
- Total number of people pass through ZoR (in order to get to FAR)