The Disguise That Didn’t Work

A series of violent robberies had shaken several rural service stations and off-licences. The latest involved an assault on staff. A witness captured a partial plate from the suspect’s vehicle: dark red saloon with a cracked rear light. The suspect wore gloves, a hoodie and a disposable mask.

Footage arrived from three systems: the shop’s internal cams, a petrol forecourt camera and a nearby council feed. Nine hours of video, one 90-second incident. We had to get an ID — fast.

This isn’t just a nice-to-have tool. It’s vital.

We loaded all footage into S21 CCTV. No need to convert — even the tricky H.264 formats worked instantly.

We started with motion analysis, which trimmed nine hours of video into four minutes of relevant action. In the clearest frame from inside the shop, we ran facial enhancement, applying shadow removal, deblurring and upscaling to expose enough facial structure — just one visible eye and an eyebrow.

That one frame gave us a match. Using facial appearance search, we tracked the same individual across other angles — outside, unmasked. Later, he appeared again in council footage, removing gloves while entering an alley.

Next, we ran vehicle detection for dark red saloons. Out of several hits, two had visible damage. One had a cracked rear light and partial plate match. We confirmed the plate using OCR as the car passed another junction.

We used motion detection to isolate the incident and remove hours of irrelevant footage. That gave us speed — and structure. Facial enhancement on the only viable frame uncovered detail invisible to the naked eye.

Once we had a viable ID, we corroborated the vehicle movement timeline using ANPR data and cross-view validation from other cameras. The full chain — entry, offence, exit, discard point — was assembled and presented for arrest. The forensic imaging unit confirmed the enhanced facial output was suitable for legal use. No footage ever left our secure system.

The process reduced the need for manual review by an estimated 20–25 staff hours and because of its offline capability, we avoided the delays and risks of cloud-based processing entirely.

One masked clip became a multi-camera timeline — with suspect, vehicle and movement all accounted for.

What if we’d dismissed the clip as too masked or grainy? What if we’d wasted time converting formats or splitting footage across systems?

Without the facial enhancement workflow, we wouldn’t have had enough detail to justify the search. Without vehicle filtering and OCR, the partial number plate might have been a dead end. Without a coherent timeline, it’s likely we’d have failed to link location, identity, and intent.

These robberies were escalating. The next one could have ended with more than an injury. In cases like this, momentum matters — you lose it, you risk losing everything.

S21 CCTV gave us what we needed: speed, precision, and proof.

Because we acted with the right tools at the right time, a pattern of violence was stopped cold.