Falls from height remain the leading cause of construction fatalities, and they are the classic serious-injury precursor: almost every fall is preceded by a visible unsafe condition that sat on camera for minutes or hours. No supervisor can watch every edge on every level at once. SafeKey watches every camera continuously and alerts your team while the hazard is still a near miss, not an investigation.
A fall takes under two seconds. The conditions that cause it last much longer. A harness worn but never clipped, a guardrail lifted out for material access, a void cover slid aside, a crew working a slab edge that was barricaded yesterday but is open today. These are not freak events. They are ordinary site conditions that stay visible, on camera, until someone walks into them.
The supervision maths never works in your favour. One safety officer covers multiple levels, rotating subcontractor crews and a site that changes shape every week. A patrol sees each edge twice a day. The scaffold gets modified at 06:30, the roof crew starts before the morning walk, and the barricade that came down for a delivery never goes back up. Between inspections, nobody is looking.
That is what makes falls preventable in a way many incidents are not. Every fall is preceded by minutes, often hours, of visible unsafe condition. Catch the precursor and you never meet the consequence. Miss it and you are reading witness statements.
Your CCTV already sees all of it. The gap has never been the cameras. It is that no human team can watch forty live feeds for the one frame that matters. That is the job SafeKey was built for.
SafeKey applies zone rules to the cameras that already overlook your height work. Draw a work-at-height zone on the camera view and set the rule: a harness is required inside it. Anyone who enters without one, or works unclipped, triggers an alert to the responsible supervisor in under a second.
The same logic covers the rest of the fall chain. Person-near-edge detection flags workers within a set distance of an unprotected edge. Barricade presence checks alert you when edge protection is removed and not reinstated. Exclusion zones under suspended loads and overhead work catch the people below who never saw the lift plan.
Rules are set per zone and per phase of work, so coverage tightens wherever the risk moves: the roof this month, the facade the next.
There is no new hardware to buy for a standard deployment. Any camera that produces an RTSP stream connects, which covers almost every CCTV system installed in the last decade. Tower crane cameras and facade or mast-mounted cameras work too, and they are often the best view you have: they look straight down on exactly the zones where height work happens.
A standard site is live within 24 hours of camera access. Zones and rules are configured with your safety team, checked against the method statement, then adjusted as the structure grows. When the risk moves from the fifth storey to the tenth, the rules move with it. No re-cabling, no site shutdown, no waiting for a hardware order.
Processing runs at the edge, fully on-premise or hybrid, depending on your data residency requirements. Footage can stay on site while alerts and dashboards reach the people who need them. Invigilo is ISO 27001 certified and the platform is built for GDPR and PDPA compliance, with no facial recognition anywhere in the pipeline.
Fixed CCTV, PTZ, tower crane and facade mast cameras all connect over a standard RTSP stream. No new hardware for standard deployments.
A standard site is operational within 24 hours of camera access, with height zones and rules configured alongside your safety team.
Processing runs where your data residency rules need it to. Footage can stay on site while your team gets the alerts and the dashboard.
WhatsApp, Microsoft Teams, Telegram and email, routed through the escalation chain you define. Unacknowledged alerts move up the chain.
HDB Singapore ran Invigilo across an active estate construction site on existing cameras, with work at height among the monitored risks. The numbers below were measured on site.
The alert is only half the value. Every detection is saved as a timestamped clip, so the near miss that used to vanish into a shift change becomes evidence. Safety teams run toolbox talks on last week's actual footage instead of generic slides, and every corrective action traces back to the exact moment it was raised.
Yes. Harness detection runs on standard CCTV footage over an RTSP stream, with no special cameras required. You mark the work-at-height zones on each camera view, and SafeKey flags any person inside a zone without a visible harness. Detection accuracy is verified at 85% and above across live deployments.
A wide view of the zone works best: one camera that sees the edge, the access point and the working area in a single frame. Most sites already have this from a standard installer layout. Any RTSP stream can be used, including tower crane and mast-mounted cameras looking down on the deck.
No. It extends their coverage to every hour and every camera at once, which no roster can do. The system raises the condition; your people keep the judgement calls about how to respond on the ground.
No. There is no facial recognition. SafeKey detects conditions, not identities: a person without a harness, a person near an edge, a barricade that is missing. This keeps monitoring compliant with GDPR and PDPA and keeps the focus on fixing conditions rather than blaming individuals.
Under a second from detection to the alert reaching WhatsApp, Microsoft Teams, Telegram or email. Fast enough for a supervisor to act while the person is still near the edge, not after the shift report.
The alert goes to the responsible supervisor with a timestamped clip attached, following the escalation chain you define. If it is not acknowledged, it escalates. The corrective action is then tracked to closure, so an open edge gets fixed the same day instead of resurfacing in next month's audit.
A proof of concept on the cameras that already overlook your riskiest edges, live within 24 hours of camera access.