4K, colour night vision and thermal: choosing a camera for the scenario

Sensor technology in security cameras has advanced a great deal in recent years. Resolutions of 4K (8 megapixels) and above are common, "colour at night" cameras show a colour image in low light, and thermal cameras have dropped in price and entered commercial projects. But the right choice starts with one question: what does the camera need to see, and at what level of detail?

Resolution and level of detail

The professional measure is pixel density on the target (pixels per metre), not megapixels alone. The international standard IEC 62676-4 describes levels such as monitoring, detection, observation, recognition and identification, each requiring a different density. A 4K camera with a very wide lens may give less detail on a face at the entrance than a 4MP camera with a suitable lens.

Sensors and low light

Sensor size, lens aperture and image processing determine low-light performance more than resolution. Colour night-vision cameras combine a sensitive sensor, a wide aperture and sometimes soft white LED lighting. The advantage: details such as the colour of a car or clothing are kept at night. The drawback: white light is noticeable, and in very dark places IR illumination is still more effective.

Dynamic range (WDR)

A building entrance with a glass door facing a bright street, or a car park with headlights, creates extreme differences in light. A camera with true WDR (based on multiple exposures) keeps detail in both the bright and the dark areas. This is one of the most important parameters at store and office entrances.

Thermal cameras

A thermal camera does not "see" light but heat radiation. It is excellent for perimeter detection in open areas, in total darkness, smoke or fog, and is not dazzled by headlights. It is not suitable for face identification or reading plates. Radiometric thermal cameras can also measure temperature and are used to monitor equipment overheating, but this is not a substitute for a standards-compliant fire detection system.

Compression and storage

Higher resolution means more data. Codecs such as H.265 and "smart" compression that adjusts bitrate to scene activity greatly reduce storage and network load. In design, disk capacity should be calculated from the number of cameras, resolution, frame rate, retention period and site activity.

A practical decision table

• Main entrance/tills: identification level, strong WDR, lens matched to distance.
• Car park: LPR at the gate, overview cameras with IR or colour night vision.
• Long perimeter fence: thermal, or cameras with analytics, combined with lighting.
• High-bay warehouse: narrower lenses along aisles, sometimes panoramic cameras.
• Server room/armory: a quality indoor camera alongside a standards-compliant alarm.

A numerical example: level of detail

Suppose you want to identify faces at an entrance four metres wide. A camera with an image width of about 2,560 pixels covering four metres gives about 640 pixels per metre, in the range usually considered suitable for identification. The same camera with a wide lens covering twenty metres gives only about 128 pixels per metre: enough to detect a person but not to identify them. That is why entrances get a dedicated, close camera rather than relying on a single panoramic one.

Fixed versus varifocal lenses

• Fixed lens: cheap and simple, but one angle. Suitable when the scenario is known.
• Motorised varifocal: lets you adjust the angle at installation and later remotely. Suitable when distances vary or you want flexibility.
• PTZ: a moving camera with optical zoom. Useful for patrolling large sites, but while it looks one way it does not see the rest, so it is combined with fixed cameras.

Common selection mistakes

• Facing strong light without suitable WDR.
• Mounting too high, filming the tops of heads instead of faces.
• IR reflecting off a nearby wall or a dirty housing.
• Retention calculated on a quiet day rather than a busy one.

Testing on site

In many projects it is worth placing a sample camera, recording by day and night, and checking in the recording (not just the live view) that the required details are kept. The difference between live view and a compressed recording can be significant.

A rough storage calculation

Suppose ten cameras, each averaging about four megabits per second, with thirty days' retention. Four megabits per second is about half a megabyte per second, roughly 43 gigabytes a day per camera, or about 1.3 terabytes a month. Ten cameras therefore need about 13 terabytes, plus a margin. The result varies a lot with the actual bitrate: motion-only recording, smart compression and lower frame rates reduce it, while high activity, rain or noisy night images increase it. Check the real bitrate after installation rather than relying on a calculator alone.

Supplementary lighting

Good night images often start with good lighting. Basic exterior lighting at the front and car park improves both the picture and deterrence, and may allow simpler cameras.

Summary

The right camera is the one that fits the scenario: distance, light, angle and required level of detail. A professional site survey, sometimes with a sample camera on site, avoids choosing by an impressive spec sheet that does not give the picture you need.