Knowledge

Low-Light and Night Vision for Video Analytics

Most security incidents happen when it is dark, which is exactly when cameras and detection are at their weakest. That is not a coincidence worth ignoring.

What actually degrades after dark

Two separate things get worse at night, and conflating them leads to buying the wrong fix.

A noisier image is not merely uglier; it actively degrades every stage downstream of it.

The main approaches, honestly compared

Infrared (IR) illumination

The camera emits near-infrared light, invisible to people, and switches to a monochrome night mode. It is cheap, ubiquitous and effective — within its range. The limitations are consistently underestimated: stated IR range assumes a reflective subject, falls off sharply with distance, and is defeated by rain, fog and snow, which scatter the light back into the lens. IR also reflects hard off nearby surfaces, so a camera mounted under a soffit or behind glass often blinds itself.

Low-light (starlight) sensors

Larger sensors with bigger photosites gather more light, producing usable colour images at illumination levels where an ordinary camera would have switched to monochrome. Colour is genuinely valuable — clothing and vehicle colour survive, which IR discards entirely. These cameras cost more and still need some ambient light; they are not magic in absolute darkness.

Thermal

Thermal cameras detect emitted heat rather than reflected light, so they are unaffected by darkness and largely by fog and rain. For detecting that a person is present across a large perimeter, nothing else is as reliable. The trade-offs are cost, low resolution, and that a thermal image cannot identify anyone or read a plate — it is excellent for detection and poor for evidence, which is why thermal is often paired with a visible camera.

The cheapest improvement is usually lighting

Before specifying a more expensive camera, consider adding white light. Well-placed lighting lets an ordinary camera stay in colour mode, keeps exposure times short enough to avoid motion blur, and has a deterrent effect no sensor provides. It is frequently the best value intervention available, and it is often skipped because it feels less technical.

Camera settings matter more at night than in daylight

Why confirmation matters more at night

Because noise and motion look alike after dark, a system that alerts on movement alone is at its least trustworthy precisely when it matters most. A design that requires confirmation — that a proposed region actually contains a person, sustained over time rather than in one frame — is far more robust to sensor noise, because grain does not persist as a coherent person-shaped object across seconds.

This is also why presence and dwell hold up better at night than raw motion: a swaying shadow does not accumulate dwell time, and a person standing still does.

Practical checklist

Where ManasaView fits

ManasaView confirms that a proposed region contains a person and that the presence is sustained, rather than alerting on pixel change — which is what keeps night-time false alarms down. Every alert carries a clip, so an operator can judge a noisy night scene themselves instead of trusting a label.

Frequently asked questions

Does night vision mean the camera sees in complete darkness?

Only thermal cameras genuinely work with no light at all. Infrared cameras supply their own near-infrared illumination, so they work in darkness but only within the range of that illuminator. Low-light or starlight sensors need some ambient light and produce colour images at levels where ordinary cameras cannot.

Why do I get more false alarms at night?

Low light forces the sensor to amplify, producing noise, and noise looks like change to motion-based detection. Headlights, moving shadows and rain add more. Systems that confirm a person is present and sustained, rather than alerting on pixel change, are far more resistant to this.

Is thermal worth the cost?

For reliably detecting that someone is present across a large or unlit perimeter, thermal is unmatched and unaffected by darkness, fog or rain. But thermal cannot identify a person or read a plate, so it is usually paired with a visible camera when evidence matters as well as detection.

Does night footage use more storage?

Often significantly more. Sensor noise does not compress well, so a variable-bitrate camera can sit near its ceiling all night while using far less in daylight. Storage should be sized on night-time bitrate rather than a daily average.

What is the single cheapest way to improve night performance?

Usually adding white lighting. It keeps the camera in colour mode, allows shorter exposures so moving subjects are not blurred, improves both detection and evidence quality, and adds deterrence — frequently for less than the cost of upgrading the camera.

← Back to Knowledge