Camera jargon, translated

Motion detection is built into almost every camera and recorder and uses very little processing power. Because it only measures change, it reacts to rain, insects, headlights, shadows, moving trees and spiders on the lens as readily as to a person. That is why motion-triggered email alerts are often switched off within a week of installation.

Rather than measuring pixel change, object detection classifies what it sees, so a rule can apply to people and ignore everything else. This cuts irrelevant alerts dramatically compared with motion detection, and it makes footage searchable by what was in it. It is still statistical rather than certain, so some detections will be wrong in both directions.

Examples include a shadow flagged as a person, or a reflection counted as a vehicle. Tuning detection thresholds, zones and schedules reduces false positives, but pushing too far in that direction starts missing real events, which are false negatives. Any honest discussion of an analytics system is about where that balance is set, not about eliminating both.

Tripwires are usually drawn along a fence line, across a gate, doorway or aisle, and can be set to trigger in one direction only. Combined with object detection, a line can be set to respond to people or vehicles and ignore everything else. They work best where the crossing point is well defined and the camera is close enough to classify what crossed.

Zones are how you stop alerts coming from a public footpath, a neighbour's driveway or a tree that moves in the wind, while keeping the area that matters active. They can also be used positively, for example a zone over a stock cage or a loading bay. Well-drawn zones usually do more to reduce noise than any other single setting.

A plate reading camera needs a tight lens, a shallow angle to the plate, a controlled distance and often its own exposure settings, which is why a general-purpose camera rarely does it well. Results depend heavily on vehicle speed, lighting and plate condition, and no system reads every plate. Stored plate data is personal information and should be treated accordingly.

Counts are commonly used for staffing decisions, trading-hour analysis and comparing one entrance with another. Accuracy depends on camera angle, with an overhead view on the entry line generally performing better than an angled one, and on how often people pass in groups. Counts are best treated as a consistent trend indicator rather than an exact tally.

In retail, dwell time near a display or queue helps show where attention and bottlenecks are. In security use, an unusually long dwell near a gate, dock or parked vehicle outside trading hours is often more meaningful than the presence itself. Like counting, it depends on the camera being able to keep track of the same subject while they remain in view.

An IP camera has its own network address and streams digital video over standard network cabling, usually powered by the same cable. It is the normal choice for new business systems because it supports higher resolutions, works with a wide range of recorders, and can be reached by software over the network. Older analogue cameras use coaxial cable to a DVR instead.

2MP is 1080p, 4MP is roughly 2560x1440 and 8MP is 4K. More pixels can mean more detail, but only if the lens, the light and the distance support it. A 4K camera in poor light over 40 metres will often give a worse usable image than a well-chosen 2MP camera closer in.

It is the most useful way to describe whether footage will be good enough for a particular job, because it combines resolution, lens and distance in one figure. Commonly quoted starting points are around 25 to 40 pixels per metre to detect a person, 100 or more to recognise someone you know, and 200 or more to identify a stranger. Treat those as rough guides, since lighting changes the result significantly.

A short focal length such as 2.8mm gives a wide view with less detail at distance. A longer focal length such as 12mm gives a narrow view with much more detail on a distant subject. Choosing focal length is how you trade coverage against detail, and it matters more to the final result than resolution alone.

Field of view follows from the focal length and the size of the sensor. A wide field of view covers more of a room or yard but spreads the available pixels thinly across it. Working out the field of view at the distance that matters is how you avoid buying a camera that sees everything and proves nothing.

Varifocal cameras let an installer set the exact view on site instead of guessing the right fixed lens beforehand. Motorised versions can be adjusted remotely, which is useful on a pole or high mount. They cost more than fixed-lens cameras, so they are usually reserved for views where getting the framing exactly right matters.

Most outdoor cameras have built-in infrared LEDs and switch to a black and white image when they turn on. Infrared has a stated range that falls off quickly, reflects badly off rain, dust and high-visibility clothing, and can wash out a face at close range. Some cameras instead use white light illumination, which keeps the image in colour and can act as a deterrent, but is visible and can annoy neighbours.

WDR combines differently exposed captures, or applies tone mapping, so a person standing in a doorway against bright sky is not reduced to a silhouette. It matters most at shopfronts, roller doors and car park ramps where the lighting contrast is extreme. Implementations vary widely in quality between models.

Low-light sensors use larger pixels and more sensitive electronics to keep recording in colour at light levels where an ordinary camera would switch to infrared. On a site with existing street, yard or dock lighting, this often produces a more useful image than infrared does, because colour detail survives. It is not the same as thermal imaging and it still needs some light to work with.

An IP rating has two digits: the first covers solid objects and dust, the second water. IP66 means dust-tight and able to withstand powerful water jets, which is the common outdoor benchmark. IK ratings cover impact resistance, with IK10 the highest commonly quoted level and relevant where vandalism is a concern. A rating describes the housing as tested, not the quality of the installation, and a poorly fitted cable gland will defeat any of them.

A PTZ can cover a wide area and zoom in on detail, which makes it useful where someone is actively watching. The limitation is that it only records where it is pointing, so anything happening elsewhere is not captured. Moving parts also wear, and grit and salt shorten their life, so several fixed cameras often prove more reliable than one PTZ.

Bullets are cylindrical, easy to aim and usually have a built-in sun hood. Domes sit in a clear bubble, look discreet and resist tampering, but the bubble attracts dust, salt haze and internal glare. Turrets combine an aimable camera in an open housing, avoiding bubble glare, and have become the common default. Fisheye cameras use a very wide lens to cover a whole room from one ceiling point, with software correcting the distortion, at the cost of detail toward the edges.

Each Australian state and territory has its own Act covering when recording is permitted, with the rules on listening devices generally stricter than those on video. Common themes include not recording private activities without consent and restrictions on where cameras may point. This is general information only, and the applicable rules depend on where the site is, so check your own jurisdiction or seek advice.

Some jurisdictions have specific workplace surveillance requirements, which commonly involve notifying staff in advance and not recording in areas such as toilets and change rooms. Separately, many businesses have obligations under the Privacy Act for the personal information they collect, including footage. This is general information rather than legal advice, so confirm what applies to your business and state.

Clear signage at entrances and in covered areas is a widely used practice and is expected or required in a number of situations. It also makes staff and customer expectations clear, which heads off complaints later. Requirements vary by state and by circumstance, so treat this as general information and check the position for your site.

PoE means one cable per camera and no power point at the mounting position, which is why it is standard for IP systems. Cameras draw different amounts of power, so the switch or recorder has to have enough total power budget for all of them, and heaters or illuminators push some cameras into higher PoE classes. Standard runs are limited to about 100 metres, beyond which an extender, a local switch or fibre is needed.

ONVIF defines common ways to discover a camera, pull its video and control basic functions. In practice support is partial: a third-party camera may record fine on another brand's NVR while losing analytics, audio or configuration features. Checking ONVIF compatibility before mixing brands avoids an unpleasant surprise at commissioning.

An RTSP address identifies a specific camera stream and is how most software, including analytics platforms, pulls live video. Cameras usually offer a high-quality main stream and a lower-bandwidth sub-stream on separate addresses. RTSP should be kept on the internal network and reached remotely through a VPN rather than exposed to the internet.

Recorders and cameras can only support a limited number of simultaneous connections before performance suffers. A local restream server pulls the stream once and distributes it to everyone watching, so the recorder is not hit by every viewer. It also makes it easier to feed a live view into analytics without adding load to the camera.

Putting cameras on their own VLAN keeps them away from office computers, so a vulnerable camera cannot be used to reach business data, and camera traffic does not interfere with everyday work. Each VLAN normally has its own subnet, which is the range of IP addresses that identifies devices on that segment. It needs a managed switch, and it is worth asking an installer whether they will do it.

It was once the standard way to get remote access, and it is now generally discouraged. A forwarded port exposes the recorder's login to automated scanning, and recorders are a well-known target with a long history of default passwords and unpatched firmware. A VPN or a vendor-managed remote connection achieves the same access without publishing the device to the internet.

With a VPN, a phone or laptop joins the site network as if it were on site, then reaches the recorder on its internal address. Nothing about the camera system is exposed to the public internet, which removes most of the risk associated with port forwarding. Many business routers and firewalls can provide one, and the practical limit on remote viewing is usually the site's upload speed.

An NVR holds the hard drives, manages recording schedules and retention, and provides search and playback. Many also power the cameras directly through built-in PoE ports. A DVR is the older equivalent for analogue cameras connected over coaxial cable, and some hybrid recorders handle both.

DVRs were standard before IP cameras became common, and plenty of Australian businesses still run one. They are limited by what analogue cabling can carry, so resolution and features are generally lower than an IP system. Existing coaxial cable can sometimes be reused for newer HD-over-coax cameras, which is worth asking about before rewiring a site.

Bitrate is what actually determines both image quality and storage use, far more than resolution on its own. A 4K camera set to a low bitrate will look blocky when something moves. Cameras can be set to constant bitrate, which uses the same data whether the scene is busy or still and makes storage easy to predict, or variable bitrate, which saves space on quiet scenes but makes retention harder to forecast.

H.264 is the safe compatibility choice: nearly all recorders, browsers and video management software can decode it. It uses more storage and bandwidth than newer standards for the same quality. Where a site mixes equipment from different eras, H.264 is often the format everything agrees on.

H.265 substantially reduces storage and bandwidth, which is why it is the default on most current cameras. The trade-off is compatibility: some older recorders, browsers and third-party software cannot decode it, and it takes more processing power to play back. Vendor variants such as smart or plus modes compress further, sometimes in ways other software handles poorly.

Higher frame rates capture fast movement more smoothly but use proportionally more storage. Many business sites record general areas at around 10 to 15 fps and reserve higher rates for tills, gates and doorways where a quick hand movement or a passing vehicle matters. Dropping frame rate is usually a better saving than dropping resolution.

Retention is the result of bitrate, frame rate, camera count and drive capacity, and most recorders simply overwrite the oldest footage once the drives fill. Many Australian businesses settle on somewhere between two weeks and three months, depending on how long it typically takes for an incident to come to light. Insurance requirements, tenancy conditions and the practical delay in noticing stock loss are the things that usually decide it.

RAID levels differ: mirroring keeps a full copy on a second drive, while parity arrangements spread redundancy across several drives and give more usable capacity. It protects against a drive failing, not against theft of the recorder, fire or accidental deletion, so it is not a backup. Surveillance-rated drives designed for continuous writing matter at least as much as the RAID choice.

Without one, a brownout or short outage can corrupt recordings, damage a drive or simply leave a gap at the worst moment. A modest UPS on the recorder and the network switch is usually one of the cheapest reliability improvements available. Sizing depends on how long you want to keep recording, and on storm-prone sites it also absorbs some of the small surges that shorten equipment life.

Event clips typically capture a period before and after the trigger, for example roughly ten seconds either side, so the lead-up is visible. They make review and sharing practical, because a clip can be attached to an email or handed to police without exporting hours of footage. Continuous recording usually still runs underneath as the authoritative record.