HIGH-RISK MANUFACTURING
Safety that intervenes before the fall, not after
Construction & Infrastructure — Elevated / Roofing Works. AI vision safety monitoring, smart fall-arrest wearables, UWB real-time location (RTLS), edge AI, and multi-sensor fusion.
BUSINESS REQUIREMENT
What the client needed
Our client — an infrastructure contractor executing a metro station roofing project — faced the hazard that dominates every elevated worksite: falls. On a work area spanning up to 100 m × 40 m, with 10–15 workers on the deck at any time, safety supervision depended entirely on human eyes.
WHAT WE FOUND
Findings on the ground
- Invisible tie-off risk: On a work area spanning up to 100 m × 40 m with 10–15 workers on the deck, whether a worker near an unprotected edge was clipped to a rated anchor was invisible — no camera can see a carabiner engage at that range, and no supervisor can watch every worker every second.
- Human-eyes-only supervision: Safety supervision depended entirely on human observation for the hazard that dominates every elevated worksite: falls.
- Temporary site constraints: The metro station roofing project was live for roughly two months before the crew moves on. Conventional fixed safety infrastructure — cabled cameras, cloud-dependent platforms, permanent installations — doesn't fit that reality.
WHAT WE DID
The engineering response
Mekosha Technologies engineered a modular, portable worksite safety stack, purpose-built around fall prevention and designed to redeploy from site to site with minimal re-setup. Three sensor layers, each matched to the hazard it can reliably carry, fused into one decision engine at the edge.
- AI vision across the whole deck: A dual-camera pod — wide-angle plus 30× PTZ — mounted centrally, monitoring the entire work area for PPE compliance (hard hat, hi-vis), person detection, and danger-zone logic. No hardware on the worker; deployable in days.
- Smart harness that verifies tie-off: An instrumented fall-arrest harness confirms actual tie-off to a rated anchor and detects an arrested fall through its onboard motion sensor. Clipped-in stops being an assumption and becomes a verified, continuously reported state.
- Centimetre-level UWB positioning: UWB real-time location gives exact worker positions — so edge-proximity rules run on real coordinates, not camera estimates, and every alert names the person rather than "someone over there."
- Hero rule: edge proximity + not clipped = immediate alarm: The fusion engine cross-checks position against tie-off state in real time. A worker approaching a defined edge without a verified anchor connection triggers a local hooter and a supervisor push alert instantly. When a fall occurs, the harness reports it to the base station in under a second and cues the PTZ to slew, capture, and record.
- Everything on-site, nothing in the cloud: All processing runs on a GPU edge module at a self-contained base station; LTE provides off-site visibility only. The system operates fully even with zero connectivity — a hard requirement for real construction environments.
OUTCOMES
What changed
Verified clipped-in status for every worker at height
previously unknowable at range.
Edge-proximity violations trigger alarms in ≤3 seconds; fall events reach the base station sub-second
Edge-proximity violations trigger alarms in ≤3 seconds; fall events reach the base station sub-second.
Every alert is tied to a specific worker identity
actionable supervision, not vague warnings.
Full-area PPE compliance monitoring across a 100 m span from a single mount
Full-area PPE compliance monitoring across a 100 m span from a single mount.
All-wireless
All-wireless, factory-calibrated stack that redeploys to the next temporary site without rebuilding from scratch.
Detection scope and known limits documented up front so site teams trust the alarms they receive
Detection scope and known limits documented up front so site teams trust the alarms they receive.
RELATED STORIES
From paper logs to a connected factory
Four SMT lines on a unified Industrial IoT platform — live OEE, QR lot-to-board traceability, station interlocks, and SAP integration.
Read the case study →
One management plane for every endpoint — OpenIY MDM security programme
OpenIY MDM as the management plane for a regulated NBFC — five platforms, provable policy enforcement, and the trust anchor for patch, identity, and Zero Trust access.
Read the case study →Bring us a line, a fleet, or an audit finding. We'll bring the architecture.
A discovery workshop with our solutions engineers — your assets on a whiteboard, no generic pitch deck.