Security Camera DORI Calculator
How far can your security camera actually identify a face? Enter the resolution and lens, and see the exact distances at which it can detect, observe, recognise and identify a person, plus what you would need to identify someone at the spot you care about.
A camera that records a person is not the same as a camera that can identify one. The difference is pixel density: how many pixels land on a person at a given distance. This DORI calculator works that out for your camera and tells you the maximum distance for each of the four levels in the international standard IEC 62676-4: detect, observe, recognise and identify. If you want to know how far a security camera can see a face clearly enough to be useful, this is the number that matters.
How to use the DORI calculator
- Choose the resolution. Select the horizontal pixel count of the stream you actually record. Many cameras record a lower-resolution substream to save storage, and that is the one that counts.
- Choose the horizontal field of view. Use the horizontal angle from the spec sheet. The lens presets are approximate, because the same focal length gives a different angle on different sensor sizes.
- Set the distance. Drag the slider to the spot you care about: the front step, the end of the driveway, the side gate.
- Read the verdict. It tells you which DORI level the camera reaches at that distance and, if it falls short, how much resolution or how narrow a lens you would need to identify someone there.
What DORI means
DORI is defined in IEC 62676-4, the international standard for designing video surveillance systems. Each level is a minimum pixel density measured across the scene at the target distance.
| Level | Pixels per metre | Pixels per foot | What you can tell |
|---|---|---|---|
| Detect | 25 | about 8 | Something, or someone, is there. |
| Observe | 62.5 | about 19 | Clothing colour, general build and what the person is doing. |
| Recognise | 125 | about 38 | Whether it is someone you have seen before. |
| Identify | 250 | about 76 | A face clear enough to identify a stranger beyond reasonable doubt. |
Each level needs roughly twice the density of the one below it, so for any given camera the identification distance is a quarter of the observation distance and a tenth of the detection distance. That is why a camera that seems to see the whole garden can still fail to produce a usable face at the gate.
Face identification distance by resolution and lens
Maximum distance at which each camera reaches identification level (250 pixels per metre), calculated with the same method as the tool above.
| Resolution | 110° (about 2.8 mm) | 90° (about 4 mm) | 55° (about 6 mm) | 42° (about 8 mm) | 28° (about 12 mm) |
|---|---|---|---|---|---|
| 1080p (1920 px) | 9 ft | 13 ft | 24 ft | 33 ft | 51 ft |
| 4MP (2560 px) | 12 ft | 17 ft | 32 ft | 44 ft | 67 ft |
| 4K (3840 px) | 18 ft | 25 ft | 48 ft | 66 ft | 101 ft |
Double each figure for recognition, multiply by four for observation and by ten for detection.
How the calculation works
At a distance d, a camera with horizontal field of view θ sees a scene 2 × d × tan(θ / 2) wide. Divide the horizontal pixel count by that width and you have the pixel density at that distance. Rearranging the same formula gives the maximum distance for any density threshold, which is how the calculator finds each DORI range.
This is the standard rectilinear model used in professional surveillance design. It assumes the person is near the centre of the frame, the lens is in focus and the lighting is good. Very wide lenses, above about 120°, are usually fisheye designs that squeeze the edges of the image, so detail there is lower than the model predicts.
Resolution or lens: which matters more?
Both scale the identification distance in a straight line, but the lens is usually the cheaper lever. Going from 1080p to 4K doubles the horizontal pixel count and so doubles the distance. Going from a 90° lens to a 42° lens roughly multiplies the distance by 2.6 at the same resolution, and a narrower lens does not quadruple your storage and bandwidth the way 4K does.
The trade-off is coverage. A narrow lens sees a narrow slice of the scene. That is why good systems split the job: a wide camera for overview and detection, and a separate narrow camera aimed at a chokepoint for identification. Asking one wide camera to do both is the most common reason homeowners have hours of footage and no usable face.
Why real footage falls short of the numbers
DORI thresholds describe ideal conditions. In practice, several things eat into them:
- Compression. Heavy compression, low bitrates and recording on a substream all throw away detail the sensor captured.
- Low light and infrared. Night footage is noisier and monochrome, which makes identification harder at the same pixel density.
- Motion blur. Slow shutter speeds at night smear a walking face.
- Viewing angle. A camera looking steeply down sees the top of a head, not a face, however many pixels it has.
- Focus, dirt and weather. A smeared dome, rain on the lens or a slightly soft focus all cost detail.
A sensible rule is to design for about 1.5 times the identification threshold at the spot that matters, so the camera still delivers on a dark, wet night.
Where to place identification cameras
Identification cameras belong at chokepoints, the places a person has to pass close to the camera and face it: the front door, the side gate, the garage service door, the path from the driveway. At the front door that job usually falls to the doorbell camera, so it is worth getting its height right with the doorbell camera height calculator and aiming it properly, as described in where your doorbell camera should actually point.
Mount identification cameras low enough to see faces head-on, typically 7 to 9 ft, and aim them along the direction people walk rather than across it. The room-by-room camera placement guide covers heights and angles for every entry point, and where break-ins actually happen helps you decide which entrances deserve an identification camera first.
Common DORI mistakes
- Buying resolution to solve a lens problem. A 4K camera with a 110° lens still only identifies faces to about 18 ft.
- Judging by the live view. The live stream on a phone looks sharp; the recorded clip is what you would hand over, and it is often lower quality.
- Using the megapixel figure. Two cameras with the same megapixels can have different horizontal pixel counts depending on the aspect ratio.
- Relying on digital zoom. Zooming into a recording enlarges pixels; it does not add any.
- Covering a whole yard with one camera and expecting it to identify people at the far fence.
If you want the numbers worked out for every camera on your property, a home security risk assessment maps each camera to the job it needs to do and the density it needs to do it.
Frequently asked questions
What does DORI stand for?
DORI stands for Detect, Observe, Recognise and Identify. It comes from IEC 62676-4, the international standard for video surveillance design, and describes four levels of detail defined by how many pixels fall on a subject per metre of scene width.
How many pixels do you need to identify a face?
The identification level in IEC 62676-4 is 250 pixels per metre of scene width, about 76 pixels per foot. Recognition needs 125 pixels per metre, observation 62.5 and detection 25. For reliable results at night or in bad weather, aim for about 1.5 times the identification figure.
How far can a 1080p security camera identify a face?
It depends on the lens. A 1080p camera with a very wide 2.8 mm lens of about 110 degrees reaches identification level at roughly 9 feet. With a 90 degree lens it is about 13 feet, and with a narrow 8 mm lens of about 42 degrees it is around 33 feet.
How far can a 4K security camera identify a face?
A 4K camera has 3,840 horizontal pixels, twice as many as 1080p, so it doubles the distance for the same lens. That is roughly 18 feet with a 110 degree lens, 25 feet at 90 degrees and about 66 feet with a 42 degree lens.
Is higher resolution or a narrower lens better for identifying faces?
Both increase identification distance in proportion, but a narrower lens is usually the cheaper improvement and does not increase storage and bandwidth the way higher resolution does. The trade-off is that a narrow lens covers less of the scene, which is why an overview camera and a separate identification camera often work best.
What is the difference between recognise and identify?
Recognition means you could confirm that a person is someone you already know, such as a neighbour or delivery driver. Identification means the face is clear enough for a stranger to be identified beyond reasonable doubt, which needs twice the pixel density of recognition.
What is the difference between pixels per foot and pixels per metre?
They measure the same thing in different units. One metre is about 3.28 feet, so divide pixels per metre by 3.28 to get pixels per foot. The identification threshold of 250 pixels per metre is about 76 pixels per foot.
Does the DORI distance change at night?
The pixel density stays the same, but usable detail drops. Night images are noisier and usually monochrome under infrared, and slower shutter speeds cause motion blur. In practice, expect to need a closer distance or more pixel density at night than the calculator shows for daytime.
Why can my camera not identify people at the distance the calculator shows?
Common causes are recording on a lower-resolution substream, heavy compression, poor lighting, motion blur, a dirty or out-of-focus lens, or a steep downward viewing angle that shows the top of the head. The calculator gives the best case; real footage needs some margin.
Does digital zoom help identify faces?
No. Digital zoom enlarges existing pixels and adds no new detail, so it cannot turn observation-level footage into identification-level footage. Only more pixels on the subject, from higher resolution, a narrower lens or a closer camera, can do that.
What lens should I use to cover a driveway?
For identifying people at the end of a driveway 30 to 50 feet away, a narrow lens of roughly 30 to 45 degrees at 4MP or higher is usually needed, or a varifocal lens set to that angle. A wide lens is fine for detecting cars and people, but not for faces at that range.
How does the calculator handle wide-angle and fisheye lenses?
It uses the standard rectilinear model, which is accurate near the centre of the frame. Lenses wider than about 120 degrees are usually fisheye designs that compress the edges, so detail there is lower than estimated. The calculator flags this when you choose a very wide angle.
What is IEC 62676-4?
IEC 62676-4 is the part of the international video surveillance standard that covers application guidelines, including how to specify the detail a camera must capture. Its DORI pixel-density levels are widely used by security designers to choose lenses and camera positions.
Where should an identification camera be mounted?
At a chokepoint where people must walk toward it and face it, such as the front door or a side gate, and low enough to see faces head-on, typically 7 to 9 feet. Aim it along the path people walk rather than across it.
Want every camera specified to the right DORI level?
Our independent consultants choose lens angles, resolutions and positions for each camera on your property, based on the job it needs to do. We sell no hardware and take no commissions.