Turret vs Bullet Cameras for a Factory Floor: Which Goes Where

On a floor where software reads PPE, machine state and idle time, the housing barely decides anything. What decides whether the model gets usable pixels is the viewing angle, the focal length and the mounting height. A turret and a bullet on the same sensor are near-identical. "Turret vs bullet" is mostly the wrong question — set the sightline first, then the shell is a wrapper around it.

Search this topic and every page gives you the same US-security-blog answer: bullets deter, turrets dodge night glare, domes resist vandals. That framing treats a camera as a thing that records and scares people. On an Indian shop floor the camera feeds an AI that has to see and understand a hard hat, a stopped conveyor, a forklift crossing a lane. Once that is the job, the deciding questions change completely — and so does the advice.

Key takeaways

• The housing is near-cosmetic. Same sensor, same lens, a turret and a bullet see almost the same picture. Angle, focal length and mount height do the real work.
• Top-down is for hard hats, not faces. A ceiling turret staring straight down reads a helmet but foreshortens the face; helmet-and-face analytics wants a near-frontal, mildly-downward angle.
• Pick the lens and height first, then the housing follows the mount: bullet on a wall, turret on a ceiling. Spec IP66 + IK10 where reachable.
• In India the lens-killers aren't in the IP rating — textile lint, metal dust, monsoon condensation and 45 °C+ shed heat are. A dome bubble fogs with lint and needs scaffolding to clean; a turret lens wipes at reach.
• From 1 April 2026, filter to the BIS/STQC-certified list first. That, plus ~20–40% import duty, decides which turret and bullet models you can even buy.

Why "turret vs bullet" is the wrong question

Strip the marketing away and a turret and a bullet with the same sensor and the same lens produce almost the same image. The shell changes how the camera mounts and how it looks to a passer-by. It does not change what the sensor sees. What actually sets coverage is three things the SERP barely names: mounting height, focal length (fixed vs varifocal), and the resulting field of view. Get those right and the housing is a detail you resolve last, by where the surface is.

The security industry even has a yardstick for "enough detail": IEC 62676-4 (DORI) — Detection, Observation, Recognition, Identification — measured in pixels per metre at the target. You need roughly 125 px/m to check a helmet (Recognition) and 250 px/m to read a face (Identification) (Axis pixel-density white paper, IEC 62676-4; see our resolution and lens guide for the full ladder). None of that number depends on whether the housing is a tube or a ball. It depends on the lens and the distance. That is the whole point.

What the AI needs to see decides the angle

Here is the take no US vendor page will give you, because their cameras only record. When a model has to read the floor, two "virtues" the SERP repeats become liabilities.

A bullet's long, narrow field of view is a counting and PPE liability. Bullets are sold for "long-range directional IR down an aisle." Point one down an aisle and people occlude each other along the axis, and heads foreshorten toward the far end. That is exactly the geometry that breaks head-counting and helmet detection. A wide, mildly-downward view across the aisle sees each person cleanly; a long telescopic shot lines them up behind one another.

A high top-down turret sees hard hats, not faces. The default advice is "ceiling turret over the open hall." Straight down, that camera reads whether a helmet is present, but it never gets a face — the head is a foreshortened disc and pixels-on-face collapse even when the px/m maths looks fine on paper. Any analytics that needs a face, or a helmet-with-worker link, wants a near-frontal, mildly-downward angle (roughly 15–30° of tilt), not a bird's-eye. So the reflexive "turret on the ceiling" can quietly starve your model. Mount for the angle the task needs first; pick the housing that bolts to that surface second.

What about IR glare and deterrence?

The consensus points are real, they are just secondary here. Camera makers report that a turret keeps its IR illuminators offset from the lens, so infrared doesn't reflect straight back into the sensor, while a dome's bubble can bounce IR into the lens once it clouds up (eufy). The mechanism is simple internal reflection: an emitter co-located behind the same glass lights up that glass. Modern domes with separated-IR rings and foam light-seal gaskets cut this a lot, so it is a design trade, not a law of the dome shape. And a bullet's obvious silhouette does deter, which earns it a place at a gate or boundary wall (Avigilon). File these under "tie-breakers," not "the decision."

The India environment kills lenses the IP rating never sees

Every US page ends the ruggedness talk at IP66 (dust and water jets) and IK10 (20-joule impact). Correct, and not the point in an Indian shed. The things that actually blind a camera here don't appear on an IP label: textile lint coating the glass, fine metal dust from grinding, monsoon condensation fogging the housing from the inside, and 45 °C+ un-airconditioned heat that derates the electronics and shortens sensor life. An IP66 dome is perfectly sealed and perfectly useless once its bubble is furred with lint.

That drives a maintenance argument no competitor makes. In a lint- or grinding-heavy shed, a dome bubble goes translucent in weeks and needs a ladder or scaffold to wipe; a turret's exposed lens is a flat surface you clean at reach. Over a year of cleaning cycles that is real labour cost, and it tilts textile (Tirupur knitwear), grinding and foundry bays toward turrets on that ground alone — before any glare argument. Walk a spinning unit at month-end and you can see it: the domes are milky, the turrets are wiped.

For genuinely wet lines, the common spec is also under-rated. Splash and hose-down is IP66/IP67; true high-pressure, high-temperature washdown in food and pharma is IP69K (ISO 20653), usually in a stainless housing. And IP/IK say nothing about explosion risk: flour dust, some metal dusts and pharma solvent areas are combustible/flammable atmospheres that legally require Ex-rated (ATEX/IECEx) enclosures, while chemical and coastal air wants corrosion-resistant stainless (C5 / NEMA 4X) and foundry heat wants a checked operating-temperature band. This ties straight back to the Factories Act, 1948 safety duties (§21 fencing of machinery, §22 work near machinery) that put cameras over those zones in the first place (India Code).

Why PTZ is a trap for a floor you must watch continuously

India factory-ops vendors love recommending PTZ for big halls. For continuous analytics it is a trap. A PTZ can only look one way at a time, so every event outside its current pan is simply unrecorded — fine for a guard chasing a target, useless for a model that must watch the whole floor every second. Two or three fixed turrets covering the hall beat one PTZ that happens to be looking elsewhere when the near-miss happens. PTZ is a control-room toy; fixed cameras are what the AI needs.

The 2026 rule that reorders the whole decision

Post-1-April-2026 the housing question is gated before you ever get to it. CCTV newly sold or installed in India must meet the Essential Requirements notified by MeitY, validated through STQC and registered with BIS. The practical effect: your housing choice is constrained by the certified-SKU list — you may simply not get your preferred shell in a compliant model. Government, critical-infrastructure and PSU jobs need full STQC certification, not just the BIS ER-01 self-certification that suffices for a private factory. Combined with roughly 20–40% import duty, the compliant field collapses toward Indian OEMs (CP Plus and peers), which changes which turret and bullet options even exist.

So the real 2026 sequence is: filter to the compliant list first, then choose by placement. On grandfathering, read it conservatively — the mandate targets new sale and installation; existing installs are not currently required to be ripped out, but that is the safe reading, not a quoted exemption, so confirm against the current MeitY/BIS notification for your site class before you rely on it.

The cost truth that makes this a low-stakes choice

Hardware is roughly 15–20% of a project's total cost; cabling, PoE-on-a-UPS, brackets, labour and the analytics do the rest. So the turret-vs-bullet agonising the SERP indulges in is economically trivial — the same sensor in either shell costs about the same, and either way it's a fifth of the bill. The only choices that move real money are camera count and the analytics use-case — which is exactly what a placement plan should optimise. Buy the housing by where the surface is, and stop thinking about it.

As a planning band (indicative, not a quote — pull written prices for your exact model and channel count): mainstream 2 MP IP bullets and turrets list around ₹2,000–₹2,700, and 4 MP around ₹3,500–₹4,500 as of mid-2026 (CP Plus listings, IndiaMART). Cheaper analog/HD units exist below that, but IP cameras start near ₹2,000. Turret and bullet of the same sensor land close, so choose by placement.

Specs at a glance - IP66 (IEC 60529) — fully dust-tight, resists powerful water jets. Hose-down and dust, not high-pressure washdown. - IK10 (IEC 62262) — housing survives a 20-joule impact, the top of the IK00–IK10 scale. - IP69K (ISO 20653) — high-pressure, high-temperature washdown (food/pharma CIP lines). - Helmet check ≈ 125 px/m; face read ≈ 250 px/m (IEC 62676-4 DORI) — set by lens and distance, not by the housing. - BIS/STQC ER from 1 Apr 2026 — buy only from the certified list; full STQC for government/critical sites.

Which goes where: the zone table

Adequate lighting is not optional either: the Factories Act, 1948, Section 17 requires sufficient and suitable lighting wherever workers pass or work (India Code). A model can only read what the sensor can see; fix the dark corner before blaming the camera.

Getting the angle right without a paid survey

Choosing per zone is really a sightline problem: which camera sees the event — a worker's head for a helmet check, a forklift crossing a lane, a machine's status light — at an angle and pixel density the model can use. Do that by eye across a 40-camera plant and you will re-aim half of them after the first night's footage comes back glared out or staring at a wall. This is the gap Mama closes: record a short phone walkthrough of the floor, and it reads the space — zones, sightlines, hazards, mounting surfaces — then returns a floor plan plus a placement plan that already says turret here at this tilt, bullet there, dome in the bay, with the analytics use-case and the pixels-on-target for each. The layout and the housing choice in a day, not after a paid survey.

FAQ

Is a turret camera better than a bullet for a factory? For the image an AI reads, a turret and a bullet on the same sensor and lens are near-identical, so neither is "better." Choose by the mount: turrets suit ceilings and open halls, bullets suit walls, gates and perimeters where their visible shape also deters. What actually decides coverage is the lens, the angle and the mounting height, not the shell.

Where should turret vs bullet cameras go on a shop floor? Bullets on walls, gates and perimeters, aimed down a line. Turrets on ceilings and overhangs over open work areas, tilted about 15–30° rather than straight down so the model gets faces and not just hard hats. Reserve IK10 vandal domes for loading bays and reachable spots that actually get knocked.

Why do people say turrets have less IR glare than domes? A turret keeps its IR illuminators offset from the lens, so infrared doesn't reflect straight back into the sensor, whereas a dome's bubble can bounce IR into the lens once it clouds up. It's an internal-reflection effect, and modern separated-IR, gasketed domes reduce it — so it's a design trade, not an inherent flaw of the dome shape.

What IP and IK rating does a shop-floor camera need? For reachable industrial positions, IP66 (dust-tight, resists water jets, IEC 60529) plus IK10 (20-joule impact, IEC 62262). Step up to IP69K for high-pressure hot washdown in food/pharma, and to Ex-rated (ATEX/IECEx) housings in combustible-dust or solvent zones — IP and IK say nothing about explosion risk.

Do new factory cameras need BIS certification in India? CCTV newly sold or installed from 1 April 2026 must meet MeitY's Essential Requirements, validated via STQC and registered with BIS. Buy from the certified-SKU list first, then choose placement; government and critical-infrastructure sites need full STQC, not just BIS ER-01. Existing installs are not currently required to be replaced, but confirm against the live notification for your site class.