Heat Stress on the Factory Floor: Can Cameras Help?

A normal camera cannot measure heat stress. It cannot read a worker's core body temperature or the WBGT heat index — both need instruments, not pixels. What an AI camera can do is flag heat-related behaviour: a collapsed worker, a line slowing in peak heat, missed water and rest breaks, or crowding at one fan.

That makes a camera the thinnest of three sensors in an honest heat plan — never the plan itself. Below is what each layer measures, why a thermal camera is worse than useless in a hot Indian shed, and the one heat number a camera can quantify that your WBGT meter cannot: throughput.

Can a camera measure heat stress?

No. The thing that actually matters — core temperature climbing while the body loses the ability to shed heat — happens inside a worker, and no optical camera sees inside.

Heat load in a workplace is assessed by the Wet Bulb Globe Temperature (WBGT) index, defined by the international standard ISO 7243:2017. WBGT is derived from natural wet-bulb, globe and dry-bulb temperatures — capturing humidity, radiant heat and air movement through those sensors — and read off a physical instrument. A camera captures none of them.

A thermal (infrared) camera reads surface temperature — skin, clothing, machinery — not core body temperature. Skin temperature is a weak, noisy proxy for heat strain, thrown off by sweat, airflow, distance and calibration.

Here is the part every FLIR-grade "fever camera" pitch leaves out, and it is decisive for India. Optical elevated-skin-temperature screening is only valid in a controlled, draft-free indoor environment of roughly 18–24°C, and needs a close, high-pixel view of the inner-canthus (tear-duct) region — a target a few millimetres wide. That is the physical opposite of a 45°C non-AC shed with fans blasting, sun through the roller door, and workers 15 metres from a wide-angle lens. In the exact heat a thermal camera is sold to monitor, its screening premise is void — not merely a "weak proxy," but out of specification. Anyone selling a plain camera as a heat-stress measurement on an Indian floor is overclaiming.

So drop the idea of measurement. The camera's real job is behavioural: the consequences and context of heat, where an ordinary RGB view with AI is genuinely strong.

What can a camera detect on a hot factory floor?

"Camera" here means a normal RGB camera with AI analytics; reliability assumes decent placement and lighting.

Read down the table and the split is stark: cameras nail the operational and behavioural signals and miss every physiological one. On our own live pilot floor, the highest-value use turned out to be the dumbest-sounding — a person-down alert reaching a phone in seconds, not minutes, so a collapse in a corner aisle isn't found twenty minutes too late. It is the same person-detection that flags a missing helmet, pointed at a worker on the floor at 46°C.

Where reliability drops: heat haze and glare off hot metal degrade detection in exactly the hottest zones; a worker who sinks down can be occluded in a packed shift; and the worst failure is false comfort — treating "the cameras watch heat" as if the hazard is handled. It watches behaviour, and only some of it.

The one heat number a camera can quantify: your output

Frame heat as a safety-only issue and you sell the factory owner nothing, because the owner buys on output. So look at the throughput row again — it is not just a safety flag, it is the productivity story, and the data is the strongest hook on this page.

• Heat could cost India up to ~4.5% of GDP by 2030 through lost labour hours (McKinsey Global Institute estimate of ~2.5–4.5%, cited in the World Bank's 2022 India cooling report).
• Around three-quarters of India's workforce — roughly 380 million people — do heat-exposed work (World Bank, 2022).
• Studies of Indian garment and foundry floors find a large share of workers already above the recommended WBGT limits in the hot season, with elevated WBGT linked to materially higher productivity-loss risk.

(Figures above are published macro estimates and should be re-verified at expert sign-off; they are directional, not audited plant data.)

Heat is already silently taxing your line — the question is only whether you can see the tax. A WBGT meter tells you the floor is dangerous; it cannot tell you what that danger costs this line today. A camera can: it is the one instrument that quantifies the heat tax as a per-line pace drop in the 2 p.m. window. That is the throughput argument no CCTV vendor and no academic page makes.

Where do cameras fit — the honest three-sensor stack

Every vendor sells you one box and calls it heat monitoring. The honest map has three sensors, each measuring something the others physically cannot:

The physiology a camera can't touch — rising core temperature, heart rate under strain — is exactly what a wearable band or patch is for; pairing heart rate with core temperature is the most accurate practical read on individual heat strain. Owning this complete map is the point: buy the box a vendor doesn't sell, and you stop mistaking one thin layer for the whole plan.

And sensors come last, not first. A heat plan built on any sensor is built upside down. The order:

1. Engineering controls — ventilation and exhaust, roof insulation or reflective "cool" roofs, spot cooling, and radiant shielding at furnaces, forging stations and dyeing vats. India-specific: evaporative (desert) coolers work in dry north-west India (Rajasthan, Gujarat) but stall in humid coastal Tamil Nadu, where the air is already near saturation — there you need mechanical ventilation, not more swamp coolers.
2. Water and rest — cool drinking water within reach of every work area; shaded or cooled rest space people can actually get to.
3. Work–rest cycling — heavy work scheduled away from peak-heat hours, rest ratios that scale with conditions, keyed to a recognised WBGT-based regime (NIOSH/ACGIH work-rest tables) rather than a guess. As a rough anchor, ISO 7243 reference WBGT limits fall to around 28°C for moderate work and ~25°C for heavy work (acclimatised) — and documented Indian foundry and garment floors routinely sit above those lines in summer. Your work-rest ratio should key off your meter's number, and that number is probably already over the line.
4. Monitoring, last — the WBGT meter, the wearable, and only then the camera, as a behavioural safety net for when the plan fails in real life: the break nobody took, the fan everyone fights over, the worker who went down.

Cameras earn that last slot for the same reason they help with other floor hazards — PPE and helmet detection or machine-guarding danger-zone detection — a continuous second set of eyes on behaviour. They answer the one question the first three controls cannot: is the plan working on the floor today, or only on paper? A written work–rest policy is worthless if the video shows the rest shelter empty at 2 p.m. in a heatwave.

Is heat monitoring required under the Factories Act, 1948?

No law mandates cameras for heat. But heat sits inside welfare duties a factory occupier already carries. The Factories Act, 1948 places worker health, safety and welfare on the occupier, with Section 13 covering ventilation and temperature and Section 18 covering drinking water — the two controls most directly relevant to heat. The exact numeric limits live in your state's Factory Rules; get them from your Factory Inspector rather than treating any single figure as a national threshold. Read the full text (India Code).

Two live updates worth confirming at sign-off: these duties are being consolidated into the Occupational Safety, Health and Working Conditions Code, 2020, which carries the same temperature, ventilation and water welfare duties forward as state rules roll out; and many Indian states now issue summer Heat Action Plans and labour-department advisories (under NDMA / NPCCHH guidance) mandating rest, water and adjusted hours in peak heat. A timestamped camera log — breaks taken, rest shelter used, a collapse detected and responded to — becomes documentary evidence that you followed those advisories. (Confirm current commencement and your state's specific advisory before relying on this.)

On the technical side, India's factory-safety body is the Directorate General, Factory Advice Service and Labour Institutes (DGFASLI) — now reconstituted as the Directorate General Occupational Safety and Health (DGOSH), though the DGFASLI name and domain remain in use. DGFASLI/DGOSH guidance plus a WBGT-based assessment is the credible backbone; camera analytics are supporting evidence, not the standard.

The practical takeaway

If a vendor says a camera "monitors heat stress," ask one question: does it measure anything (it doesn't) or flag behaviour (it can)? Build the heat plan first — ventilation, water, shade, sane work–rest scheduling. Add a WBGT meter to measure the environment and a wearable to catch the physiology. Then add a camera, cheap to bolt on once the real controls exist, to see whether the plan is holding on a 46°C afternoon — and to put a number on what the heat is costing your line.

FAQ

Can a camera measure a worker's body temperature or heat stress? No. No optical camera reads core body temperature, and a thermal camera only reads skin/surface temperature — a weak proxy whose screening premise is valid only in an 18–24°C draft-free room, i.e. the opposite of a hot factory floor. True assessment uses the WBGT index on an instrument. A camera flags behaviour around heat; it cannot measure the strain.

Are thermal cameras useful for heat stress at all? Rarely on a factory floor. Elevated-skin-temperature screening needs a cool, still, controlled environment and a close view of the tear-duct region — impossible on a wide, hot, breezy shop-floor shot. Surface temperature isn't core temperature, so it's a rough alert at best, never a diagnosis or a medical device.

Is heat-stress camera monitoring required under the Factories Act, 1948? No law mandates cameras. But the Act makes the occupier responsible for welfare, with duties on ventilation and temperature (Section 13) and drinking water (Section 18), now being consolidated into the OSH Code, 2020. Cameras don't discharge those duties — engineering controls and water do — but they can log whether your controls are actually used.

So should we skip cameras and just buy fans and coolers? Buy the ventilation, cooling, shade, water and work–rest scheduling first — that is the plan. Add a WBGT meter and a wearable to measure it. Cameras come last, as a behavioural safety net and the one tool that quantifies the heat's hit to output. Cameras on top of a good plan help; cameras instead of one are dangerous.