The incident happened at 11 p.m. You have the footage: a person, clearly a person, walking the length of the compound and stopping at the third trailer. Someone who saw them leave says they wore a red jacket and got into a blue van.
Your footage cannot confirm either of those things. Not because the camera was cheap, out of focus, or pointed wrong. Because after dark, that camera stopped recording colour at all.
This surprises almost every owner we walk a site with, because nobody explains it at the quote stage and the daytime demo looks superb.
What "night mode" is actually doing
An ordinary camera sensor is more sensitive than your eye. It responds not only to visible light, but to a band of infrared just past the red end of what you can see. That extra sensitivity is a problem during the day and an asset at night, and the camera resolves the conflict with a physical part: the IR cut filter.
In daylight the filter sits in front of the sensor and blocks infrared. That is what makes daytime colour look correct — foliage green rather than washed pale, skin tones natural, a navy coat navy rather than muddy. Without the filter, infrared leaks into all three colour channels and the colour information stops meaning anything.
When light falls below a threshold, the camera swings that filter out of the way. Removing it is the single largest sensitivity gain available to the camera, and it happens in one mechanical step — many cameras make an audible click. Now infrared reaches the sensor, and the camera can produce a usable picture in light your eye would call darkness.
But the colour channels are now being fed by light that carries no reliable colour meaning. So the camera does the honest thing and throws colour away, outputting monochrome. That grey image is not a fault or a degraded mode. It is the correct output of a sensor that has deliberately traded colour for sensitivity.
That trade is the whole subject of this article. "Day/night switching" — the phrase on your spec sheet — is the name of that trade. Once you know what is being traded, the decision in front of you becomes a real one instead of a feature you scrolled past.
Why the loss of colour matters more than owners expect
In practice, colour does three jobs that monochrome cannot.
It gives you a description that means something. "A person in a light-coloured hooded top" is worth far less than "a person in a red hooded top." One narrows a set. The other mostly does not.
It links cameras and events together. Someone crosses your yard at 11:04 and passes a second camera at 11:09. In colour, matching the two is often trivial. In monochrome, a red jacket and a dark green jacket can render as near-identical mid-greys, and now you are guessing whether that was the same person.
It carries vehicles. Vehicle colour is one of the few attributes a non-specialist can state with confidence and that other people can confirm independently. Monochrome night footage frequently cannot tell you whether a van was white, silver or pale blue.
Set that against what monochrome still does well, because it is not nothing. Presence, count, timing, direction of travel, whether a gate was opened, whether a second person waited by the fence. On a great many cameras that is the entire requirement, and colour would be a wasted expense.
The mistake is not having monochrome cameras. The mistake is having monochrome at the two or three positions where colour was the point.
IR illuminators, and the problems they create
Removing the cut filter only helps if some infrared is present in the scene. On a dark site there is little, so most cameras carry their own illuminator: a ring of infrared LEDs around the lens, invisible to you, obvious to the sensor.
Putting the light source in the same housing as the lens is convenient and cheap. It also causes a specific family of problems that show up only at night, which is exactly when nobody is looking at the live view.
Reflection off the housing. In a dome camera the lens looks out through a curved bubble. If the illuminator's light can reach the inside of that bubble — through a gap in the internal shroud, off a smudge, or past a foam gasket compressed with age — the camera photographs its own light. The result is a bloom of white haze across part of the frame, invisible during the day and ruinous every night. The same illuminator in a turret housing, where lens and LEDs sit in an open ball with no shared window, does not have this failure mode at all. Housing choice is usually sold as an aesthetic and vandal-resistance question. At night it is an optical one.
Near-object hotspots. Infrared falls off sharply with distance. Anything close to the camera — a wall return, a downpipe, a sign, a bollard, a branch — takes the full force of the illuminator and blows out white, while the camera's exposure pulls down to compensate and everything further away goes black. A camera mounted tight against a corner can spend every night correctly exposing a wall.
Spiders and weather. Warm housings attract insects, insects attract spiders, and a web strung across the front of a camera reflects infrared beautifully. Rain and snow do the same: each drop near the lens lights up, wrecking the image and generating motion alerts all night. Sites with no false-alert problem in summer discover one in November.
Retroreflection. Some materials — high-visibility clothing, road signage, and the retroreflective sheeting on vehicle registration plates — throw infrared straight back at the source with startling efficiency. This is why a plate that reads perfectly at noon can be a solid white rectangle at midnight.
None of these are reasons to avoid infrared. They are reasons to look at your own night footage rather than the brochure's.
White light: the option that keeps colour
The straightforward way to keep colour after dark is to light the scene with light the sensor can see properly — ordinary white light. The camera keeps its cut filter in place, keeps producing colour, and behaves at night more or less as it does by day.
Some cameras carry white LEDs in the housing for exactly this. Others simply rely on the site's existing lighting, which is frequently the better answer and costs nothing.
The costs of white light are not technical, which is why they get overlooked in a specification and then cause trouble after handover:
- Neighbours and light trespass. A floodlight pointed across a property line at 2 a.m. is a complaint waiting to happen. Municipalities differ on what they permit, so check locally before mounting anything that flashes on at movement.
- Occupants and tenants. Light through a bedroom or hotel-room window is a tenancy problem, not a security one.
- It announces the camera. Sometimes that is the goal — a light that snaps on is a genuine deterrent and often ends the incident. Sometimes it is precisely wrong, because you wanted to observe rather than to interrupt.
- Wildlife and moving vegetation trigger it, and a light that fires forty times a night gets disabled by whoever has to live under it.
A common middle path is white light that stays off until an event, triggered by the camera's own analytics rather than a crude motion sensor, so the site stays dark until there is a reason not to be. You still get colour for the part of the recording that matters, and you avoid lighting an empty yard all night.
Low-light colour sensors: the third path, honestly described
There is a class of camera built to hold colour in far less light than an ordinary model — larger sensors, wider-aperture lenses, and processing tuned to pull a colour image out of very little.
Where there is some ambient light — a street lamp, a lit loading bay, spill from a neighbouring building, a well-lit forecourt — these cameras can keep colour where an ordinary camera would have gone grey. That is a real capability and on the right site it is the cleanest answer available, because it requires no new lighting and creates no nuisance.
Two limits deserve stating plainly.
They are not zero-light cameras. Something has to be there to collect. In a genuinely dark rural yard with no lighting at all, no sensor produces colour, and a vendor implying otherwise should be asked to demonstrate it on your site at night.
Colour at low light is bought with exposure time or with noise, and both cost you detail. Holding the shutter open longer collects more light and smears anything moving; suppressing noise aggressively smooths away the fine texture that made a face recognizable. You can end up with a confidently coloured image of a person you still cannot identify — which is the same disappointment as the pixel-density problem, arriving by a different route.
The one honest test is the boring one: footage of a real person, at night, in the specific spot that matters, from the camera being proposed.
And thermal is not the answer to this
Thermal cameras come up in this conversation because they are the one technology genuinely indifferent to darkness. Axis, describing its own product, puts the reason plainly: "Since objects themselves emit the heat that thermal cameras detect, thermal cameras are not dependent on visible light and can detect in all light conditions."
But thermal solves the opposite half of the problem. It removes your dependence on light and removes identification along with it — no colour, no face, no plate, no logo. If your complaint is "I cannot tell who that was," thermal makes it worse, not better. It has its own article on this site, and the distinction is worth reading before anyone quotes you one.
The decision, made per camera
This is not a site-wide choice. It is a short list.
Walk your cameras and put each one in one of two buckets.
Cameras that only need to establish presence. Perimeter runs, yard overviews, corridors, stairwells, general awareness of a compound. Infrared is correct here, it is cheap, it is unobtrusive, and monochrome costs you nothing you were going to use. Most of your cameras belong in this bucket.
Cameras where you would need to describe a person or a vehicle. The main entrance. The till or reception. The gate vehicles pass through. The one blind corner where things actually happen. Usually three or four positions on an ordinary commercial site, not thirty. These are the positions where you decide, deliberately, between existing site lighting, added white light, or a low-light colour camera — and where you accept a nuisance cost or a hardware cost on purpose.
Then ask four questions of whoever is quoting you:
- Which of these cameras produce colour at night, and at what light level do they give up? Not "does it have night vision." Every camera has night vision.
- Where is the illumination coming from at each position — the site's own lighting, the camera's illuminator, or nothing?
- Which of these cameras will lose colour first as autumn comes in? A demo in July and a demo in November are different demonstrations.
- Can you show me night footage from this exact position? The recurring answer to nearly every camera question, because it is the only one that cannot be argued with.
The part we cannot promise
Even a well-designed night installation gives you less than the daytime version of itself. That is a property of there being less light, and no equipment removes it. A camera that holds colour at dusk may still hand you monochrome at 3 a.m. in December rain, and a white light that keeps colour on your forecourt does nothing for the dark ground beyond it.
What you can have is a system where the loss is chosen — where you know which cameras go grey, you decided that was acceptable at those positions, and the two or three that mattered were built to hold colour. That is a different situation from discovering the trade after an incident, which is how almost everyone currently discovers it.
