Capability
CCTV face recognition — the difference between 'someone is there' and 'who is there'
CCTV face recognition is what turns a motion alert into a decision. Instead of telling you a person is at the gate, PGAK tells you whether it's your shift supervisor, a delivery driver you've seen forty times, or someone the system has never seen before — and only the last one is worth waking you up for.
How cctv face recognition works
Enrolment
Each person who belongs — staff, family, regular contractors — is enrolled from a handful of frames. It takes seconds per person and can be done from existing footage.
Template, not photograph
The face is converted into a mathematical vector and the image is discarded. There is no searchable photo library of your employees sitting on a disk.
Matching at the edge
Every face the cameras see is vectorised and compared against the enrolled set, on hardware at your site. Nothing is sent to an external service to do this.
Act on the result
Known face during expected hours: silence. Unknown face at a sensitive door: alert with a snapshot. Known face somewhere they shouldn't be: alert too.
Where it earns its keep
- Gate attendance at factories, replacing fingerprint machines that fail on dusty hands
- Silent entry for family in a home system, so notifications stay worth reading
- Unknown-adult alerts near school gates during school hours
- Restricted-room access logs in hospitals and offices
- Repeat-visitor detection near high-value retail shelves
What it won’t do
Every AI vendor lists strengths. These are the limits, so you can plan around them instead of discovering them.
- A face that is more than about 60° from the camera, heavily covered, or lit only from behind will be detected as a person but may not be identified.
- Recognition quality depends on camera placement — a camera at gate height facing arrivals will always outperform one mounted high in a corner.
- It identifies enrolled people. It cannot tell you the name of someone who has never been enrolled, and no honest system claims otherwise.
Face recognition — common questions
Can face recognition work on my existing CCTV cameras?
Yes, in most cases. What matters is placement and resolution at the point of recognition rather than the camera being marketed as 'AI'. A standard 2MP camera at gate height facing arrivals works well; a 4K camera mounted high on a corner often doesn't.
Are photos of my staff stored somewhere?
No. Faces are stored as mathematical templates that cannot be reversed into a usable photograph, and processing happens on a device at your premises rather than in a cloud.
How accurate is it?
On a well-placed camera with a frontal view, matching an enrolled person is highly reliable. Accuracy drops with extreme angles, heavy backlighting or covered faces — which is why we treat camera placement as part of the deployment rather than an afterthought.
Does it work with masks or helmets?
Partially. A mask covering the lower face reduces confidence significantly and a full helmet usually prevents identification. In those environments we lean on person detection, zones and schedules rather than on identity.