Learn / Surveillance
DORI explained: pixel density for security cameras
What detect, observe, recognize and identify mean, how to calculate them, and why paper numbers can be optimistic.
Security camera specifications often say a camera can "identify at 20 m" or "detect at 80 m". These claims come from DORI: detect, observe, recognize and identify, defined in the standard IEC 62676-4 by pixel density on the target. Here is what each level means, how to calculate it from a lens and sensor, and why a camera that meets the number on paper can still fall short.
Pixel density, not resolution
DORI levels are defined in pixels per meter (px/m) at the target: how many sensor pixels cover one meter of the scene at that distance. A 4K camera with a wide lens can have lower pixel density on a person than a 1080p camera with a narrow lens. Resolution alone does not tell you what you can see; the lens and distance matter just as much.
| Level | Pixel density | Pixels on a 1.7 m person | What you can tell |
|---|---|---|---|
| Monitor | 12.5 px/m | ≈ 21 | Crowd flow and general activity |
| Detect | 25 px/m | ≈ 43 | Whether a person is present |
| Observe | 62.5 px/m | ≈ 106 | Clothing and what someone is doing |
| Recognize | 125 px/m | ≈ 213 | Whether it is someone you know |
| Identify | 250 px/m | ≈ 425 | Who someone is, beyond reasonable doubt |
| Inspect | 1000 px/m | ≈ 1700 | Small details such as plates or documents |
The one formula you need
For a rectilinear lens with focal length f, pixel pitch p and target distance D, each pixel sees a patch of p · D / f on the target, so
pixel density = f / (p · D)
with all lengths in meters. Rearranged, the maximum distance for a level is D = f / (p · PPM), and the focal length needed to reach a level at a given distance is f = PPM · p · D. Note that the pixel count does not appear: for a given pixel pitch, density depends only on focal length and distance. More pixels give a wider scene at the same density, not more detail on the target.
A worked example
A 1080p camera with 2.9 µm pixels and a 4 mm lens:
- At 10 m:
0.004 / (2.9×10⁻⁶ × 10) ≈ 138 px/m, which meets Recognize but not Identify. A 1.7 m person is about 234 pixels tall. - Identify (250 px/m) works out to about 5.5 m, Recognize 11 m, Observe 22 m and Detect 55 m.
- To identify at 10 m, the lens needs
250 × 2.9×10⁻⁶ × 10 ≈ 7.3 mm, which narrows the horizontal field of view from about 70° to about 42°.
See pixel density, DORI distances and a pixelated target for your own setup.
Pixels-on-target calculator →Why paper DORI can be optimistic
Pixel density is a necessary condition, not a guarantee. Real images lose detail in several ways the formula ignores:
- Lens sharpness. Pixels only count if the lens resolves them. A lens with low MTF at the sensor's Nyquist frequency, or one stopped down into diffraction, blurs detail across several pixels.
- Off-axis targets. The formula holds near the image center. Toward the edges, distortion and the oblique view reduce pixel density, especially with wide lenses.
- Compression. Video encoders discard fine texture first, and faces are exactly that.
- Low light and motion. Noise reduction smears detail, and long exposures blur moving people.
- Focus. A fixed-focus lens set for a near distance may be slightly soft at the far DORI range, or the reverse.
A sensible practice is to design for the next level up, or to keep a margin of 25–50% above the threshold at the distances that matter.
Check whether diffraction limits detail at your pixel pitch and f-number.
Diffraction & resolution →Key points
- DORI levels are pixel densities on the target: 25, 62.5, 125 and 250 px/m.
- Pixel density is f / (p · D); the sensor's pixel count sets coverage, not detail.
- Maximum distance for a level is f / (p · PPM).
- Lens MTF, compression, light and motion all eat into paper DORI, so keep margin.