Skip to content

DORI and Pixel Density at Distance: The Complete CCTV Design Guide

Resolution alone is a meaningless surveillance specification. A 4K camera can fail to identify a subject at 12 metres while a lower-resolution unit succeeds — the difference is governed entirely by pixel density at distance, written PD(d). This guide derives the PD(d) and field-of-view equations from first principles, maps them to the four DORI tiers defined in IEC 62676-4, and shows how to specify the exact lens needed to satisfy any tier at any range.

Open the free CCTV Design Tool →

1. The governing standard: IEC/EN 62676-4 and the DORI model

DORI — Detection, Observation, Recognition, Identification — is the framework defined in IEC 62676-4 (and BS EN 62676-4) for stating, objectively, what a camera image is good enough to do at a given distance. Each tier corresponds to a minimum pixel density across the target, expressed in pixels per metre (px/m).

Tier Minimum pixel density What it permits
Detection25 px/mA person is present in the scene
Observation63 px/mSome characteristic detail, clothing, action
Recognition125 px/mRecognise a known individual with high certainty
Identification250 px/mEstablish identity beyond reasonable doubt

Why "1080p" tells you nothing on its own

A camera's resolution only matters relative to how wide a scene those pixels are spread across at the target distance. The same 1920-pixel-wide sensor delivers Identification-grade density at 4 m and barely Detection at 40 m. The specification that matters is always px/m at the distance you care about.

2. Deriving pixel density at distance, PD(d)

2.1 The scene-width equation

The horizontal width of the scene a camera sees at distance d is set by its horizontal field of view (HFOV):

W(d) = 2 · d · tan(HFOV / 2)  [metres]

2.2 The core formula

Pixel density is simply the horizontal pixel count spread across that width:

PD(d) = H_res / W(d) = H_res / ( 2 · d · tan(HFOV / 2) )  [px/m]

2.3 Worked example

Take a sensor with H_res = 1920 px, a lens giving HFOV = 90°, and a target at d = 15 m. The scene width is W = 2 · 15 · tan(45°) = 30 m, so PD(15) = 1920 / 30 = 64 px/m — just past the Observation tier (63 px/m), but far short of Recognition. To recognise a face here you would need a narrower lens or higher resolution.

3. The lens–sensor relationship behind HFOV

3.1 Focal length is the real lever

HFOV is fixed by sensor width and focal length:

HFOV = 2 · arctan( sensor_width / (2 · f) )

3.2 Sensor format matters

The same focal length produces a different HFOV on a 1/2.8" sensor than on a 1/1.8" sensor, because the physical sensor width differs. Always pair a focal length with its sensor format before trusting a coverage figure.

3.3 Inverting the problem

In practice you work backwards: fix the px/m you need and the distance, solve for the required HFOV, then for the focal length. The CCTV Design Tool performs this calculation from the configured camera resolution, sensor, and lens data, then renders the resulting DORI bands on a calibrated plan.

4. From pixel density to maximum effective range

Rearranging the core formula gives the maximum distance at which a chosen DORI tier (target density PDt) is met:

d_max = H_res / ( 2 · PDₜ · tan(HFOV / 2) )

Real range is shorter than this nominal figure: edge-of-frame lens distortion, low light, IR cut-in, compression artefacts, and motion blur all erode usable density. Treat d_max as a ceiling, not a guarantee, and design with margin.

Stop calculating by hand. The free CCTV Design Tool runs this exact maths on every camera, draws the four DORI bands on your floor plan, and flags blind spots automatically.

Launch the free tool →

5. Applying PD(d) in real layout engineering

Zone your site into Detection, Recognition, and Identification bands — an entrance lobby needs Identification (250 px/m) at the door line, while a perimeter only needs Detection. Place each camera so its tier band lands exactly where the requirement falls, validate the layout against your floor plan, and export an auditable report. For number-plate work, the same model applies: licence-plate Recognition (ANPR/LPR) requires roughly 250 px/m across the plate, scored per vehicle approach angle.