How night vision cameras work: infrared vs starlight vs thermal, what lux ratings mean, IR range, colour night vision, and getting reliable footage at night.
Night vision is a goal reached three ways: infrared adds invisible light and gives a black and white picture; starlight and low light cameras make the most of existing light and often show colour; thermal sees heat and needs no light at all.
The most important night specification is minimum illumination in lux. Insist on the colour figure, not the infrared one.
At night, sensor size and lens brightness often matter more than megapixels.
Infrared range, not resolution, is often the real limit after dark, and fog, rain and nearby reflective surfaces all work against it.
Thermal detects; night vision identifies. Serious sites use both, and Visylix brings infrared, starlight and thermal cameras together on one platform.
A night vision camera is any camera built to keep producing a usable picture as light fades, from dusk into full darkness. Night vision is a goal, not one technology, and manufacturers reach it in three ways that are easy to confuse.
Infrared cameras bring their own invisible light. Starlight and low light cameras make the most of the light already there. Thermal cameras build a picture from heat and need no light at all. Choosing well starts with knowing how each behaves, because each fails in different conditions.
Most affordable night vision cameras use infrared. A ring of infrared lights sits around the lens, usually at 850nm or 940nm, just beyond what the eye can see. At night the camera switches modes: it moves a filter out of the way so its sensor can see infrared, turns the lights on, and records the reflected light. The result is the familiar black and white night picture.
Two details matter to a buyer. 850nm lights are more efficient and reach further, but give off a faint red glow that reveals the camera. 940nm lights are truly invisible but reach less far. Choose based on whether you need range or discretion.
The stated infrared distance, say thirty metres, is how far the lights usefully reach. Beyond it the picture goes dark, so infrared range, not resolution, is often the real limit at night.
Infrared has weaknesses. It only lights what it can see directly, heavy fog, rain or snow scatter it, and reflective surfaces close to the camera can flare and wash out the picture.
Starlight (a marketing name, not a standard) and low light cameras take the opposite approach. Instead of adding light, they gather more of what is already there using a large, sensitive sensor, a bright lens and strong noise reduction. With any ambient light, from the moon, streetlights or a nearby building, they can produce a colour picture where an infrared camera shows only black and white.
The key figure is minimum illumination, measured in lux, where lower is better. For reference: full daylight is tens of thousands of lux, twilight about ten, a full moon about 0.1, and a moonless starlit night roughly 0.001. A camera rated 0.001 lux in colour claims true starlight performance; one rated 0.01 lux with infrared on is a more modest low light camera. Read carefully, because vendors often quote their best case with infrared on, in black and white, rather than the colour figure buyers care about.
Colour night vision is a related feature. Some cameras keep colour at night by pairing a sensitive sensor with a small warm white light instead of infrared. That gives colour detail useful for identification, at the cost of being visible and attracting insects and attention.
Thermal is often lumped in with night vision, so it is worth being clear. A thermal camera does not use light at all; it sees the heat every object gives off, so it works in total darkness, through smoke and at long range. But it cannot show a recognisable face, read a plate or show colour. Its job is detection, not identification.
Night vision is the opposite: it needs a little light or infrared, but it shows recognisable detail. The rule is simple. To identify who someone is, use night vision. To detect that something is there across a large or pitch dark area, use thermal. Serious perimeters use both, and our thermal cameras guide covers thermal in depth.
A handful of specifications decide night performance.
Minimum illumination in lux: the single most important night figure. Insist on the colour value.
Sensor size: larger sensors gather more light, so they see better at night than smaller ones of the same resolution.
Lens brightness: a lower f number lets in more light.
Infrared range and wavelength: match them to the area you need to cover, and choose 940nm if the camera must stay discreet.
Resolution, with care: squeezing more megapixels onto a small sensor makes each pixel smaller and worse in low light, so at night a lower resolution camera with a bigger sensor often wins.
Wide dynamic range: helps with the mix of bright lights and deep shadow common at night.
Every night vision camera is still an IP camera, so the networking, PoE and protocol basics of any IP camera apply too.
The most common mistake is treating the camera as the whole answer and ignoring the scene. A few proven principles help.
Add a little light where you can. A modest amount of ambient light transforms low light colour cameras and can remove the need for infrared altogether.
Avoid infrared flare. Do not aim an infrared camera at a nearby wall, sign or window; the reflection whites out the picture. Mount it so the light spreads into open space.
Mind the weather. Fog, heavy rain and snow scatter infrared, so in those climates a starlight camera near existing light, or thermal for detection, is more dependable.
Plan for identification. Keep the area where you need to recognise faces or read plates within a lit band and at a sensible distance. Detection range and identification range are very different numbers.
Combine camera types. A PTZ camera with strong infrared for zooming in, fixed low light cameras for context and thermal for long range detection is a common, effective mix.
A frequent question is how well AI such as person detection or line crossing works at night. AI works on whatever picture the camera delivers, so its reliability follows picture quality. A clean, well lit low light picture supports AI almost as well as daylight; a noisy, dark, washed out one weakens every analytic that relies on it.
So the fastest way to improve night time AI is to improve the night time picture: a better sensor, sensible lighting, correct infrared and careful mounting.
Strong AI helps too. Visylix detects small, distant subjects that many systems miss, and every camera learns its own normal, including its night time scene, so motion and anomaly alerts reflect what is unusual for that camera after dark. Camera tampering detection also flags infrared failure, so a camera that has quietly lost its night vision is caught before an incident, not after.
To a video platform, a night vision camera is just another IP camera sending a stream. Visylix adds infrared, starlight and thermal cameras over ONVIF and RTSP alongside the rest of your estate, records them on your own servers, and runs its 22 AI analytics, all built in house, on those feeds: intrusion detection, line crossing, person tracking, fire and smoke and more.
Because Visylix works with cameras from any major brand, mixing infrared cameras on the perimeter, low light cameras at identification points and thermal for long range detection is simply a matter of adding cameras, not building integrations. ONVIF device packs cover the top five camera brands, and bulk administration makes large estates quick to set up.
When something happens at night, Visylix acts: Sudarshan rules alert the right people, raise alarm incidents in the operator console and can talk down an intruder through the camera speaker, while Radha, the AI copilot, lets an operator ask "show me everyone near the east fence after midnight" in plain language.
Starter and Pro include unlimited cameras on the VMS licence, from ₹4,999 or $49 per month, so layering several camera types across a site never multiplies your licence. AI analytics are added per camera per month on the cameras you choose.
Most security cameras use infrared. A ring of infrared lights floods the scene with light the eye cannot see, at 850nm or 940nm, the camera removes a filter so its sensor can see that light, and it records a black and white image. Starlight and low light cameras instead use a large, sensitive sensor to make the most of whatever light is already there, often producing a colour image.
An infrared camera brings its own invisible light and produces a black and white image, limited by how far its lights reach. A starlight camera adds no light; it relies on a sensitive sensor and a bright lens to see with the light already present, and can often show colour at night. Infrared works in total darkness within range; starlight needs a little ambient light but shows more detail.
Lux measures how much light falls on a scene. A camera's minimum illumination rating is the lowest light level at which it still produces a usable picture, and lower is better. A full moon gives roughly 0.1 lux and a moonless starlit night about 0.001 lux, so a camera rated 0.001 lux in colour is claiming genuine starlight performance. Always check whether the figure is for colour or for infrared black and white.
Infrared cameras can, within the reach of their own lights. Starlight and low light cameras cannot, because they need some light to work with. Thermal cameras can, because they see heat rather than light.
Neither is better; they do different jobs. Thermal detects presence in total darkness and at long range but cannot identify a person. Night vision can identify people but needs some light or infrared. Choose thermal to detect, night vision to identify, and both for demanding perimeters.
Yes. Colour at night needs either enough ambient light for a sensitive sensor or a small warm white light built into the camera. Pure infrared night vision is always black and white, because infrared carries no colour.
They should, if they support open standards. A night vision camera is a normal IP camera, so a video platform that supports ONVIF (the open camera standard) and RTSP, such as Visylix, can add, record and analyse it alongside daytime cameras with no special integration.