Facial recognition in video surveillance

Xeqmate facial recognition turns cameras into face detection points: every person who walks past becomes a searchable face read, feeding live monitoring, face alerts and the persons-of-interest file — with permission controls, an audit trail and privacy governance built into the platform from the start.

Detection on camera or server Real-time face alerts Permission-based access Privacy by design
Concept

What is facial recognition in video surveillance?

Facial recognition in video surveillance is the automatic detection and comparison of faces in camera footage. Every detected face becomes a face read, with a photo, date, time and location, which can be searched, compared against registered photos and used to raise alerts when a person of interest is recognised. At Xeqmate, detection runs on the camera itself (when it has AI on board) or on the platform server, from the video of any RTSP/RTMP camera.

In practice, facial answers questions raw video cannot: has this person been here before? When, and through which cameras? Are these two faces the same person? Has the person on the alert list just walked in? The answers land in the Event Center and in the operation notification channels — and, because biometric data is involved, all of it happens under permissions, auditing and controlled retention, as detailed in the responsibility and privacy law section.

Features

What the Xeqmate facial module does

Real-time monitoring

Face detections appear live as people pass the cameras — the control room follow-along screen.

Face read search

Search the history by period, camera and attributes — including by image: upload a photo and find similar reads.

One-to-one face comparison

Upload two photos and see the similarity score between the faces — direct support for identity checks in investigations.

Face alerts

Register the photo of a person of interest, pick the cameras and the trigger schedule, and get the notice when they are recognised — over push, email, Telegram and alert groups.

Face flow report

Detection volume per camera and period — the quantitative view of people movement at the monitored points.

Detected attributes

Reads can indicate sex, age range, skin tone, mask use and a liveness check — useful filters for narrowing a search.

Investigation

The persons-of-interest module

Facial recognition gains depth when it is tied to a structured file on the people being monitored.

A file per person

A record with photos, names and aliases, documents, organisations and tags — and face-alert creation straight from the file, pre-filled with the primary face.

Relationship graph

A view of the relationships between people, organisations and incidents — the connections a flat list will never show.

Person map

People, points of interest and areas of activity on the map — the geographic context of the investigation.

Points of interest and areas of activity

Places and regions linked to each person or organisation, organising where to look and what to monitor.

Responsibility

Biometric data is sensitive data — and the platform treats it that way

Privacy laws treat biometric data as a special category: GDPR Article 9 in Europe, LGPD in Brazil, and state statutes such as Illinois BIPA, Texas CUBI and Washington’s HB 1493 in the United States. Using facial recognition takes more than technology: it takes governance.

  • Lawful basis and purpose: processing faces requires a defined lawful basis and a specific purpose — it is the customer who establishes them, and the platform is designed to operate inside them, with transparency and respect for data-subject rights.
  • Permission-based access: the facial module only appears for authorised users; each one sees only the sites, cameras and modules the administrator releases.
  • Audit trail: queries, alerts and administrative actions are all logged — the operation knows who searched which face, and when.
  • Controlled retention: the retention days for face reads are set by the customer, separately from video retention; anything past the window is discarded.
  • Two-factor authentication: 2FA available for users, with active-session management.

Facial recognition is a tool for operations with defined accountability — central stations, security companies and public safety — and it does not replace the customer duty to handle biometric data under the privacy law that applies to them.

Integrations

Faces integrated with public-safety networks

Smart Sampa

Two-way sharing with the City of São Paulo municipal network (Brazil) — cameras, plates and facial recognition integrated.

Muralha Paulista

Sending faces and plates to the São Paulo State Government programme (Brazil), with alerts of interest.

Face alerts also reach the customer own system through a signed webhook, and reads can be queried through the external API — facial is one of the modules of the Xeqmate VMS SaaS, alongside LPR and the video analytics.

FAQ

Questions about facial recognition

Not necessarily. Detection can run on the camera itself (models with face AI on board) or on the platform server — in which case any camera with an RTSP/RTMP stream will do, as long as it delivers the face large, frontal and lit. Mounting at eye level, a narrow crossing point and light on the face matter more than the camera brand.

Beyond the face itself, reads can indicate sex, age range, skin tone, mask use and a liveness check. Those attributes work as filters when searching reads.

You upload the face photo, fill in a name or alias, a category and a description, then pick the cameras that trigger and the schedule (days and hours). When the person is recognised, the notification arrives through the configured channels — email, Telegram and up to 5 alert groups — and the event shows up in the Event Center. Alerts expire on the set date (90 days by default) and can be renewed.

It depends on where you operate. In the United States there is no single federal biometric statute, but state laws such as Illinois BIPA, Texas CUBI and Washington HB 1493 require notice and, in several cases, written consent before collecting face templates — with private rights of action in Illinois. Xeqmate supports that governance with permission-based access, an audit trail, customer-controlled retention and 2FA; defining the lawful basis and the purpose is the responsibility of whoever operates the system.

A structured record of the people being monitored: a file with photos, names, documents, organisations and tags; a relationship graph; a person map with points of interest and areas of activity. From the file, a face alert is created pre-filled with the primary face.

Most cases come down to three causes: backlight (the face turns into a silhouette), a camera mounted too high (you only see foreheads and caps), and a face too small in frame. Mount close to eye level, at a narrow crossing point, with light on the face — and use frontal, sharp, recent photos when registering alerts.

Live demo

See facial recognition on one of your cameras

30 minutes with our team: we connect one of your cameras live, you watch the face detections arrive in real time, and you leave with the cost for your own scenario.

We reply within one business day · no credit card, no install, no commitment