Types of security camera, and which suits which job

What types of security camera are there, and which one suits which part of your site?

Most commercial sites need more than one type of security camera. Fixed cameras cover the doors, gates and counters you need evidence from. PTZ or multisensor cameras cover wide areas where something needs following. Thermal cameras detect people where there is no light, and licence plate recognition cameras record vehicles at the gate. The right type for each position comes from one question, asked before any model is chosen: does this camera need to detect someone, recognise them, or identify them?

Start with the job, not the camera

The video surveillance standard AS/NZS 62676.4:2020, the New Zealand and Australian adoption of IEC 62676-4, sets what a camera position needs to deliver, measured in pixels per metre of the scene. The more detail a position needs, the narrower the area one camera can cover at that standard. That is why one camera cannot do every job, however many megapixels it has.

DETAIL LEVELS IN AS/NZS 62676.4

Detect: 25 px/m	You can see that a person or vehicle is there.
Observe: 63 px/m	You can see what they are doing, for example at a loading bay or across a yard.
Recognise: 125 px/m	You can tell whether it is someone you know.
Identify: 250 px/m	You can identify someone you do not know. This is the level evidence usually needs.

A camera that covers a whole car park at detection level will not identify anyone in it. Doors, gates and counters, where people pass close and slowly, are where identification-level cameras earn their cost. The same logic applies to perimeters. On construction sites, the standard offer in this market is a camera on a pole with a solar panel at the front gate. It watches the entrance and leaves the rest of the boundary open, which is how Faisandier Group’s development sites were covered before its cameras and detectors were moved to follow the whole boundary.

The main camera types, and where each falls short

• Fixed cameras (bullet, dome and turret). One view, recorded all the time. The right choice for doors, gates, counters, loading docks and production lines. They see only what they were pointed at, and a lens set for a wide view will not identify anyone at the far end of it. The shape is mostly about mounting and deterrence: a bullet camera is obvious, and a dome is harder to tamper with and harder to read the direction of.
• PTZ cameras (pan, tilt and zoom). One camera that can be steered and zoomed across a large area, by an operator or automatically. Good for yards, perimeters and car parks where something needs following. While it looks one way, it is not recording the other, so a PTZ adds to fixed coverage and does not replace it.
• Multisensor and panoramic cameras. Several sensors in one housing, covering 180 or 360 degrees without gaps. Good for yards, aisle intersections and open-plan floors. Detail drops with distance, so they observe well and identify poorly.
• Thermal cameras. They see heat rather than light, so they detect people and vehicles in full darkness, smoke and light fog, and shadows and headlights do not fool them. Good for perimeters and large dark areas, usually paired with analytics. They cannot identify anyone: thermal tells you something is there, and another camera has to tell you who.
• Licence plate recognition (LPR) cameras. Built to read plates on moving vehicles, with fast shutters and infrared to cut headlight glare. Good at gates, weighbridges and car park entries. Angle and speed matter more than resolution here, so a general camera pointed at a gate rarely reads plates reliably.
• AI-enabled cameras. Any of the types above with analytics on board: telling a person from a vehicle or an animal, line crossing, loitering and counting. Worth having where nobody watches live and the system has to decide what deserves an alert. Accuracy depends on placement and set-up, and some features are sold harder than they perform. Our article on AI camera analytics covers what to ask a supplier to demonstrate.
• Vandal-resistant and hazardous-area cameras. Vandal-resistant housings are rated for impact under IEC 62262, where IK10 is the highest common rating. Hazardous-area cameras are certified for explosive atmospheres, such as fuel, solvent or dust environments, under the AS/NZS 60079 series. If part of your site has a hazardous area classification, a standard camera in that zone is a compliance problem, not just an equipment choice.

A real estate usually combines several of these. Carterton District Council’s 7 sites run multi-head, PTZ, wide-angle and licence plate recognition cameras on one platform, with the LPR camera installed specifically to support Police investigation work.

What New Zealand privacy law expects of a camera system

Cameras that record people collect personal information, so the Privacy Act 2020 applies. The Privacy Commissioner’s CCTV guidance sets out what that means in practice:

• A purpose, written down, before the cameras go in. Position each camera to meet that purpose without unreasonable intrusion. Bathrooms, and views into neighbouring property or through windows, are out.
• Signs where people will see them before they are recorded. Put them near the cameras and at the edge of the area covered, naming who operates the system and how to contact them.
• A retention period. Keep footage only as long as the purpose needs, then delete it. Our article on recording methods covers how to set one.
• Access for the people recorded. People can ask to see footage of themselves. Have a way to show it while protecting anyone else in the same footage, and log who outside the business has had access, including Police.
• Security. Protect footage from loss, unauthorised access and alteration.

Facial recognition adds a further set of rules. The Biometric Processing Privacy Code came into force on 3 November 2025, and the grace period for organisations already using biometrics ended on 3 August 2026. It covers automated processing that verifies, identifies or categorises people. Before collecting biometric information, a business has to be satisfied that it is necessary, effective and proportionate for a clear purpose, put safeguards in place, and tell people it is happening. Ordinary recording falls outside the Code. Switching on face matching brings a system inside it. The Privacy Commissioner’s 2025 inquiry into a supermarket facial recognition trial is the clearest public guide to what that takes in practice.

Licence plate records need the same care. A plate tied to a time and a place can be personal information about the driver, so apply the same purpose, retention and access rules to LPR data as to footage.

Check your cameras against their jobs

1. List every camera and write down its job: detect, observe, recognise or identify.
2. At each door, gate and counter, pull a recent clip of a person and ask whether you could identify a stranger from it. If you could not, that position is under-specified for its job.
3. Check that cameras cover the exits you would need evidence from, not just the entrances.
4. Walk the site and check the signs: near the cameras, at the edge of coverage, with the operator’s name and contact details.
5. If any camera runs face matching, check that you have a written proportionality assessment and a notice that meets the Biometric Processing Privacy Code.
6. Check where alerts from thermal or AI cameras go, and who responds to them.

How Blue Eagle handles it

We specify cameras position by position, against the job each one has to do, and write that down so the design can be checked later. Where analytics are worth having, they are commissioned at installation and switched on at the level you need now. Where the cameras you already have can do the job with better positioning or set-up, we will say so.