Guide
Camera pixel density: detect, observe, recognise, identify
Short answer: pixel density is how many pixels of the image fall on one metre (or foot) of the scene. To identify a person you need about 250 pixels per metre (76 per foot) where they stand; to detect someone, 25 (7.6 per foot).
The levels (IEC 62676-4)
| Level | Pixels per metre | Pixels per foot | What you can do |
|---|---|---|---|
| Detect | 25 | 7.6 | See that someone is there |
| Observe | 62 | 19 | See what they are doing, clothing colours |
| Recognise | 125 | 38 | Recognise someone you know |
| Identify | 250 | 76 | Identify a stranger beyond reasonable doubt |
| Inspect | 1,000 | 305 | Read small details, e.g. bank notes |
The formula
Width of the scene at distance d = 2 × d × tan(field of view ÷ 2). Pixel density = horizontal resolution ÷ that width. Turned around: the distance where a level ends = resolution ÷ (2 × level × tan(field of view ÷ 2)).
Example
A 4 MP camera (2,560 pixels wide) with a 100° lens. To identify (250 px/m): 2,560 ÷ (2 × 250 × tan 50°) = 2,560 ÷ 596 = 4.3 m (14.1 ft). To recognise (125 px/m): 8.6 m (28.2 ft). Zoom the same camera to 40° and identify reaches about 14 m.
What changes the result
- A narrower lens or higher resolution reaches further – but covers less width.
- Wide dynamic range, light, compression and focus matter: pixel density is the minimum, not a guarantee.
- Fisheye and panoramic cameras spread their pixels over a much wider angle.
Questions
Is 4K always better?
Only if the pixels land on the area that matters. A well-placed 4 MP camera at a door often beats a 4K camera covering the whole lobby.
Do I need identify everywhere?
No – plan identify at entrances, cash points and chokepoints; observe or detect for general areas and perimeters.
Related
Let GITECHY Design do the math
Cable lengths, PoE and power budgets, batteries, voltage drop and camera coverage – on every design. One flat price per company.