Why PPE Detection Is Harder Outdoors Than in the Demo

PPE detection is sold as one capability and is really four, with genuinely different difficulty. Vendors demonstrate the easy one indoors and imply the rest follow. They do not, and knowing the ranking before you specify is what stops a deployment from disappointing in its second month.

The four items, ranked by how well they work

The practical consequence is that helmet and vest compliance can be monitored across a whole site from existing cameras, while glove compliance realistically needs a camera positioned at a specific task — a workbench, a chemical decant point — rather than a yard view. A vendor quoting one accuracy figure across all four items has not measured them separately, which is the thing worth asking about.

What Gulf conditions actually do to it

Overhead sun and hard shadow

Midday sun close to vertical produces the worst case for a helmet check: the brim shades the face into near-black while the hat itself blows out to white. Dynamic range, not resolution, is the limiting factor, and a camera metering for a bright yard renders the person as a silhouette. This is a camera specification and positioning problem rather than a model problem, and the fix is exposure and angle, not a better detector.

Dust on the lens and in the air

Airborne dust reduces contrast across the whole frame, which is recoverable. Dust accumulating on the housing is not: it softens every frame permanently and it happens gradually enough that nobody notices. Accuracy declining over months on a site that has not changed is almost always this, and the remedy is a cleaning schedule rather than a support ticket. It is the single most common cause of a system that "used to work".

Heat and the equipment itself

Enclosures in direct sun run hot enough that processing hardware throttles or restarts, and a camera at thermal limits produces noisier images. This shows up as detection quality varying by time of day in a way that looks like a software fault and is a shading and ventilation issue.

What workers actually wear

In summer heat a lot gets adapted: a cloth wrapped under a hard hat, a vest worn open or tied at the waist, sleeves modified, a shemagh under the helmet. Almost none of this appears in the datasets these models were trained on. A vest worn open is the common false positive for non-compliance, and a system that flags it repeatedly while the worker is in fact wearing the vest will lose the confidence of the site quickly — and rightly.

Where the camera has to be

PPE checking wants a view of a person from roughly chest height at moderate range, near-frontal, which is not how site CCTV is usually mounted. A camera high on a mast looking down at a yard sees the top of a helmet — adequate for helmet presence, poor for a vest, useless for gloves.

The approach that works is a choke point rather than area coverage. One camera at the gate or the entrance to a work area, at head height, checking people as they pass, will outperform six cameras covering the whole site — because it inspects everyone once under controlled geometry instead of inspecting everyone badly all day. It also produces far fewer alerts, which matters more than it sounds.