Quick answer: Mount face-capture cameras at 8 to 9 ft (2.4 to 2.7 m), tilted no more than about 30 degrees down, aimed where people face the camera 8 to 15 ft (2.4 to 4.6 m) away. Mount overview cameras at 12 to 20 ft (3.7 to 6 m) and accept that they identify nobody. Plates need a dedicated per-project LPR camera at a shallow angle; deterrence turrets go at 9 to 12 ft (2.7 to 3.7 m) facing the approach path. Height is half the decision; the other half is pixel density on the target and the lighting the camera is pointed into.
Which of the four goals is this camera for?
A camera that tries to do everything does nothing in court. Assign each position one primary goal and design height, angle and lens for it. The pixel-density targets below come from IEC 62676-4, the operational-requirements standard used across the industry.
| Goal | Pixel density on target | Typical height | Down-tilt | Lens |
|---|---|---|---|---|
| Identify a stranger’s face | About 76 px/ft (250 px/m) | 8 to 9 ft (2.4 to 2.7 m) | 15 to 30 degrees | 2.8 mm or 4 mm, aimed at the approach |
| Recognize a known person | About 38 px/ft (125 px/m) | 8 to 10 ft (2.4 to 3 m) | 20 to 30 degrees | 2.8 mm or 4 mm |
| Observe or detect (overview) | About 19 px/ft (62.5 px/m) to observe; 7.6 px/ft (25 px/m) to detect | 12 to 20 ft (3.7 to 6 m) | 30 to 45 degrees | 2.8 mm or a 180° unit |
| Deter (light and audio) | Classification range, not face density | 9 to 12 ft (2.7 to 3.7 m) | 15 to 30 degrees | 2.8 mm facing the approach |
Entrances get a dedicated low, shallow face camera; everything else gets an overview camera high enough to be out of reach and see over vehicles, and the customer is told in writing which is which.
How high and how steep for a 2.8 mm or 4 mm lens?
Pixel density is arithmetic: horizontal pixels divided by scene width at the target distance, where scene width is 2 × distance × tan(half the horizontal field of view). The table assumes about 100 degrees for a 2.8 mm lens and about 80 degrees for a 4 mm lens on a typical 1/2.8 in sensor. The catalog pages for the turrets below print resolution but no field-of-view row, so take the exact angle from the spec sheet and run it through the field of view and lens calculator before you promise a distance.
| Horizontal pixels (as printed) | Example SKU | 10 ft (3 m) | 15 ft (4.6 m) | 20 ft (6 m) | 30 ft (9 m) |
|---|---|---|---|---|---|
| 3840 (3840×2160) | CS-I8XQ II-2.8mm | 161 px/ft | 107 px/ft | 80 px/ft | 54 px/ft |
| 3072 (3072×1728) | CS-I5XQ II-2.8mm | 129 px/ft | 86 px/ft | 64 px/ft | 43 px/ft |
| 2688 (2688×1520) | CS-I4SV S-2.8mm | 113 px/ft | 75 px/ft | 56 px/ft | 38 px/ft |
Against the 76 px/ft identification line, a 2.8 mm lens at the assumed angle identifies to roughly 20 ft (6 m) on 8MP, 17 ft (5 m) on 5MP and 15 ft (4.6 m) on 4MP; all three recognize a known person to 30 ft (9 m). A 4 mm lens narrows the scene to about 34 ft (10 m) at 20 ft (6 m), pushing 8MP identification reach to about 30 ft (9 m) at the cost of width.
Tilt is the other half. A camera at 9 ft (2.7 m) aimed at a face 5.5 ft (1.7 m) above the floor and 10 ft (3 m) away looks down at 19 degrees; at 6 ft (1.8 m) away the angle is 30 degrees, where brows, caps and hoods start hiding eyes.
What are the face-capture rules at an entrance?
- Aim where people face the camera. Inside the door looking at people entering, or on the approach where they look toward the handle. A camera that sees the back of heads leaving is an overview camera, however good the lens.
- Choke point, not open floor. Every entrant should pass through a 4 to 6 ft (1.2 to 1.8 m) wide zone at the design distance.
- Height 8 to 9 ft (2.4 to 2.7 m). Out of casual reach, low enough to keep tilt under 30 degrees. Above 10 ft (3 m) you are recording scalps.
- Handle the glass. A door with daylight behind it silhouettes faces. Specify wide dynamic range and mount to one side of the door axis. The CS-I8XQ II-2.8mm and CS-I5XQ II-2.8mm print 120dB True WDR (TWDR); the CS-I4SV S-2.8mm prints 120dB.
- Confirm analytics separately. The gen-2 turrets print a Face Detection / Capture row (up to 32 faces / 16 faces). Enabling it depends on use case, firmware, region and legal requirements (in Canada: PIPEDA and provincial privacy law, Quebec the most restrictive). Treat it as a project-review item; see Facial recognition: when not to promise it.
Both gen-2 turrets also print a built-in microphone and speaker. If audio recording is enabled, state recording-consent rules apply in the US (in Canada: PIPEDA and provincial law); verify with counsel.
Where do plates and parking cameras go?
An overview camera on a 15 to 20 ft (4.6 to 6 m) pole covers rows and shows what happened. It does not read plates at night: headlights, retroreflective plates and a steep angle defeat a general-purpose sensor. Plate reading is a separate camera with its own geometry: horizontal and vertical angles to the lane of 30 degrees or less, a height that keeps the plate in frame across the capture zone, and a lens long enough to put the plate at the required density. Those conditions, plus vehicle speed, lighting and region, decide legibility.
CASE’s LPR camera, the CSP-I4LQ I-LPR-8-32mm, is a Project Series item: project-quoted, with availability, certificates and lead time confirmed in writing, and performance stated only per project against the spec sheet and site conditions. Capture the approach geometry on the survey and send it to Anabon; the layout method is in Parking lot security cameras.
- Overview: pole-mounted at 15 to 20 ft (4.6 to 6 m), tilted 30 to 45 degrees.
- Plates: a dedicated camera per lane, shallow angle, per-project quote.
Which lighting mistakes ruin otherwise good placement?
Most unusable night footage comes from a correctly mounted camera pointed at the wrong light.
- Under the luminaire. A camera beside or below a wall pack sees the fixture’s glare and insects, and its exposure drops to protect the highlight. Mount 3 ft (1 m) or more from the fixture and keep it out of frame.
- Into the sunrise or sunset. An east- or west-facing camera is blind for an hour a day; offset the aim or accept the gap in writing.
- IR bounce. IR LEDs reflecting off a soffit, a white wall or a dome bubble wash the image out. Keep the housing clear of surfaces inside about 3 ft (1 m), and on domes check the bubble for reflections before you leave.
- Backlit doors and docks. Specify WDR and mount off-axis.
- Relying on IR for faces. IR gives a monochrome image; clothing colors are lost. Where color at night is required, specify it per SKU: the CS-I4SV S-2.8mm prints 0.0001 Lux, 33 ft (10 m) of white light and no IR, with PhotonX; the gen-2 turrets print 0.002 Lux, 164 ft (50 m) IR and 98 ft (30 m) white light.
Low-light and full-color behavior is stated per SKU on the specification sheet and should be confirmed for the site. Where site lighting is being designed at the same time, IES recommended practices give the illuminance targets; coordinate camera positions with the lighting plan.
Where does an EW deterrence turret belong?
EW and AEW are deterrence features. Availability is confirmed per SKU. The laser is a deterrent, not an illuminator, and is offered on selected models only. On the gen-2 turrets, deterrence means the printed 98 ft (30 m) white light, red and blue light (printed Yes) and the built-in speaker, triggered by human or vehicle classification in zones you draw; trigger options are confirmed per model and firmware.
- Height 9 to 12 ft (2.7 to 3.7 m). Out of reach, yet low enough that the white light lands at eye level and the speaker is intelligible.
- Face the approach path, not the street. Aim along the route from the property line to the asset, so the light is seen early.
- Draw zones inside the property line. A zone that includes a public sidewalk fires all night and gets switched off by week two.
- Schedule it. Deterrence runs after hours; in business hours the camera records normally.
- IR for the record, white light for the effect. The CS-I8XQ II-2.8mm prints 164 ft (50 m) IR and 98 ft (30 m) white light; the record extends beyond the deterrent.
False-alarm control and zone design are in Active deterrence cameras: Early Warning (EW/AEW) light and audio vs recording only.
What to tell your customer
“Each camera has one job. The entrance cameras are mounted low and shallow to capture faces as people come in; the high cameras give an overview of the lot and are not meant to identify anyone. Each position is on the drawing with its purpose written beside it, so you know what the system can and cannot show you.”
FAQ
What is the best height for a security camera at a door?
Eight to nine feet (2.4 to 2.7 m) for face capture: out of casual reach, yet low enough to keep the down-tilt under 30 degrees at a target 8 to 15 ft (2.4 to 4.6 m) away. Above 10 ft (3 m) the camera records the tops of heads. Aim it where people face the camera.
How many pixels per foot do I need to identify a face?
About 76 px/ft (250 px/m) on the target to identify a stranger and about 38 px/ft (125 px/m) to recognize a known person, following IEC 62676-4. Divide the horizontal pixel count by the scene width at the target distance. An 8MP sensor with a 2.8 mm lens at about 100 degrees of horizontal view reaches identification density to roughly 20 ft (6 m).
Can an overview camera on a pole also read license plates?
Not reliably. Plates need a dedicated camera at a horizontal and vertical angle of 30 degrees or less, a lens that puts the plate at the required density, and control of headlight glare. CASE quotes its LPR camera per project, with performance stated against the spec sheet and site conditions; capture the approach geometry on the survey.
How much down-tilt is too much for a face camera?
Past about 30 degrees, brows, caps and hoods hide the eyes and the image stops being an identification. A camera at 9 ft (2.7 m) looking at a face 5.5 ft (1.7 m) above the floor sits at 19 degrees when the person is 10 ft (3 m) away and 30 degrees at 6 ft (1.8 m); set the capture point farther out.
Where should a deterrence turret with white light and a speaker be mounted?
At 9 to 12 ft (2.7 to 3.7 m), facing the approach path from the property line to the asset, with zones drawn inside the property and a schedule for after hours. EW and AEW are deterrence features and availability is confirmed per SKU; the CS-I8XQ II-2.8mm prints 98 ft (30 m) of white light and 164 ft (50 m) of IR.
