REC 00:00:00
CAM 01 – Parkade
Most commercial camera systems are sold on megapixel count and installed by eye. The result is a system that records everything and identifies nobody — useful for knowing that something happened, useless for proving who did it.
IEC 62676-4 defines four levels of what a camera can deliver, measured in pixels across a metre of the scene. It is the only spec that predicts whether an image will stand up.
| Level | Pixel density | What you get |
|---|---|---|
| Detection | 25 px/m | Something is there. No detail. |
| Observation | 62.5 px/m | Clothing, general behaviour, direction. |
| Recognition | 125 px/m | You can tell it is someone you already know. |
| Identification | 250 px/m | A stranger can be identified from the image alone. |
A camera that reboots at night is almost always a power problem, not a camera problem. Heaters, wipers, IR illuminators and PTZ motors all draw at once in February.
| Standard | Budget at the switch | Available at the camera | Typical use |
|---|---|---|---|
| 802.3af (PoE) | 15.4 W | 12.95 W | Fixed dome or bullet, no heater |
| 802.3at (PoE+) | 30 W | 25.5 W | Outdoor housing with heater and IR |
| 802.3bt Type 3 | 60 W | 51 W | PTZ, multi-sensor |
| 802.3bt Type 4 | 100 W | 71 W | Heated PTZ, illuminator arrays |
Switch budget is the total across all ports, not per port — a 24-port switch rated 370 W cannot run 24 PoE+ cameras. Ethernet is also limited to a 100 m channel; longer runs need fibre or a mid-span injector.
Published lux ratings are measured at settings nobody uses — slow shutter, high gain, aperture wide open. A camera rated 0.01 lux may be unusable at that light level once the shutter is fast enough to stop a walking person. Specify by sensor size and aperture, and treat quoted IR range as a detection figure rather than an identification one.
H.265 typically halves the bitrate of H.264 at equivalent visual quality, and smart-codec modes cut it further on static scenes. Both choices cascade into storage, switch uplinks and any remote viewing over a business internet connection — which is where most systems actually fall over.
Retention you can actually prove, on hardware built for it
Read more →
Credentials that cannot be copied in a parking lot
Read more →
Alarms that mean something, so someone still answers them
Read more →
Protection on day one, before there is power or a fence
Read more →
Someone watching, not just something recording
Read more →It is the wrong question. Start from what each area has to deliver — identification at doors and tills, recognition in aisles, detection across a yard — and the count falls out of the pixel-density calculation. Warehouses are usually over-covered on the floor and under-covered at the dock doors, which is where the losses happen.
Only where the extra pixels land on something that matters. Over a wide yard, 4K buys real detail. On a corridor already at 250 px/m it buys storage cost and nothing else. The decision is made per camera position, not per system.
Often, if they are ONVIF Profile S conformant. Age matters less than conformance and whether they still deliver the pixel density their position needs. We audit what is there before proposing replacement.
Each location page covers what is specific to that area — the industrial districts, the site types and the constraints that change the design.