How Existing CCTV Becomes Your Smartest Theft & Material-Loss Defence on Unattended Sites

The fix for site theft is not more guards or more cameras. It is making the cameras you already have specific enough to raise an alarm while the intruder is still on site — and specific enough to tell you whether you are looking at a genuine breach or a stray dog crossing the compound.

The loss you cannot attribute

Copper, cable, steel, fuel, formwork and small plant walk off sites every week, and most of it is discovered at the next stock check or when a machine will not start — long after the vehicle has gone. Construction carries one of the highest material-loss exposures of any sector: high-value, portable, resaleable stock sitting on an open site that is unmanned for two-thirds of every week. Traditional security — a night guard, a gate register and footage nobody reviews until something is already missing — catches a small slice of what is actually happening.

What AI vision changes

Computer vision identifies observable events that precede a loss: a vehicle entering after hours, a person crossing the perimeter line, movement in the material lay-down yard or fuel-bowser area at 2am, loitering near the store cabin. Xorth surfaces these in real time and alerts your security team or site manager while the intruder is still on site and the material is still recoverable. Detection is behaviour- and event-based — it flags an action in a zone at a time, never a person, and every alert is verified by a human before anyone responds.

The out-of-hours window is the hard half

Nights, weekends and shutdown periods are when the site is emptiest and the loss is largest, and they are exactly when nobody is watching the monitors. Time-and-zone rules turn that on its head: any motion in the lay-down yard, fuel store or plant park outside working hours becomes an instant push alert with the clip attached, so one remote operator can cover a dozen sites instead of one guard covering one gate. ANPR at the entrance logs every vehicle in and out with a timestamp and plate — which turns “a truck came in the night the cable went missing” from a suspicion into a searchable record.

Beyond theft: safety and site intelligence

The same cameras detect fire and smoke, unauthorised entry into live-plant or excavation zones, and after-hours trespass that carries a liability risk if someone is hurt on your site. They also produce gate-flow, headcount and zone-activity data off hardware you have already paid for — no separate access-control installation, no second vendor.

No new hardware

The historic barrier to smarter site security was capital cost. Xorth runs on existing IP CCTV over RTSP through your NVR or DVR; inference runs on a GPU server in the site cabin or a private cloud tenancy you control. Your security team gets specific alerts on their phones. During the survey we map models to camera positions — a floodlit gate camera is not the same asset as a dome over the lay-down yard, and only one of them will support ANPR.

Curious what your site cameras would catch after dark? We run a 30-day pilot on a single site, measured against your own incident and loss baseline, before anyone talks about scaling.