field note
How far can a security camera see? The DORI distances
How far a security camera can see depends on its pixels and its lens. DORI distances for 1080p, 4MP, and 4K cameras, and how to find the lens a spot needs.
2026-10-25
A security camera can see as far as its pixels reach: a 1080p camera with a 90 degree lens can identify a person to about 13 feet and detect one to about 126 feet, and a 4K camera with the same lens doubles both. A datasheet has no single range because “see” covers four different jobs, and each job needs a different number of pixels on the target.
What “see” means
IEC 62676-4, the international standard for video surveillance (CCTV) systems, defines four levels of detail as pixel density on the target. They are the DORI levels:
| Level | Pixels per meter | Pixels per foot | What it lets you do |
|---|---|---|---|
| Detect | 25 | about 8 | Tell that a person or vehicle is there |
| Observe | 63 | about 19 | See what a person is doing and what they are wearing |
| Recognize | 125 | about 38 | Tell whether this is someone you have seen before |
| Identify | 250 | about 76 | Identify a person beyond reasonable doubt |
Security camera placement covers which level each part of a site needs. This post is about distance: how far out each level holds, and how to work backward from a spot to the lens that reaches it.
The distance comes from two numbers
Pixel density falls with distance because the sensor’s pixels are spread across a wider scene the farther out you look. The width of the scene at any distance is set by the horizontal field of view, so two numbers fix the range at every level: the sensor’s width in pixels and the lens angle.
distance = sensor pixel width ÷ (pixels per meter × 2 × tan(field of view ÷ 2))
Worked out for three common sensors at a wide and a narrow lens setting:
| Camera | Field of view | Identify | Recognize | Observe | Detect |
|---|---|---|---|---|---|
| 1080p (1,920 px wide) | 90° | 13 ft | 25 ft | 50 ft | 126 ft |
| 1080p (1,920 px wide) | 30° | 47 ft | 94 ft | 187 ft | 470 ft |
| 4MP (2,688 px wide) | 90° | 18 ft | 35 ft | 70 ft | 176 ft |
| 4MP (2,688 px wide) | 30° | 66 ft | 132 ft | 261 ft | 658 ft |
| 4K (3,840 px wide) | 90° | 25 ft | 50 ft | 100 ft | 252 ft |
| 4K (3,840 px wide) | 30° | 94 ft | 188 ft | 373 ft | 940 ft |
Two rules fall out of the table. Double the sensor’s pixel width and every distance doubles. Narrow the lens from 90 to 30 degrees and every distance grows by about 3.7 times, while the scene at any given distance becomes about a quarter as wide. Reach and width trade against each other, and the lens setting is where the trade gets made.
What shortens it on site
The table is geometry, and geometry is the best case. Real images land short of it for reasons the formula leaves out:
- Light. At night the range is set by the illuminator as much as the sensor. A datasheet’s infrared range is a separate figure, and at a narrow lens it is often shorter than the daytime identify distance.
- Compression. A low bitrate smears the fine detail that recognize and identify depend on, most of all on a moving subject.
- Motion blur. A long exposure in low light blurs a person walking through the scene.
- Lens sharpness. A wide lens resolves less at the edges of the image than at the center.
- Mounting height. A camera 12 feet up sees a face at an angle. The distance to the face is longer than the distance across the floor, and the face is foreshortened.
- Backlight and weather. A bright doorway behind a person, rain, fog, and glare off wet pavement all cost detail.
Leave margin at identify, where a short fall means the image cannot do the one job it was installed for.
Working backward from the spot
On a site walk the question usually runs the other way. There is a gate 60 feet out and the job needs identify there: which lens? Solve the same formula for the angle. At 60 feet, identify needs a field of view of about 46 degrees or narrower on a 4K camera, about 33 degrees on a 4MP camera, and about 24 degrees on a 1080p camera. A varifocal camera fits if that angle falls between its wide and tele ends. A fixed lens fits only if its one angle is close to it.
Reach a point on the coverage map
The MentatNOC camera coverage map does that solve on the floor plan. Select a camera’s pin, choose Reach a point, and click the spot it has to see. The plan needs a scale and the camera’s model needs a lens profile, and the button says which is missing when either is. The answer gives the distance to the spot and the field of view the camera needs to hold its reach level there: 60 feet away, 45.6 degrees to identify. It uses the pin’s own reach level, so set the level first to ask about recognize or detect.
If the model can get there, the result adds the zoom position, from 0 at the wide end to 1 at full tele, and the focal length that matches it. Apply sets the camera’s field of view to that value, and the cone narrows to reach the spot. If the model cannot get there, the result says it is out of range for that model and lists the three models in the catalog whose zoom range comes closest to the angle needed. That angle was worked out at the selected camera’s resolution, so check a listed model’s sensor before swapping: a camera with fewer pixels across needs a narrower view than the one shown to hold the same level.
The distance is measured flat across the plan. A pin with a mount height and tilt draws its cone as a band on the floor, with a dead zone under the camera and a far edge shorter than the flat reach, so check the cone after applying.
Two other tools answer the related questions. Inspect works in the other direction: click any spot inside a cone and read the distance, the pixels per foot, and the DORI level that camera achieves there. A PTZ model draws a dashed ring at its full-zoom reach behind its home cone, which shows how far it could see if it turned that way. The ring never counts as coverage, because a PTZ only covers where it is pointed.
Doing this across 500 cameras
A fleet’s range question is rarely about one camera. It is a refresh across a campus, every door at identify and every lot at detect, with a mix of models already on the wall. By hand, that is a spreadsheet of sensor widths and lens angles and a calculation per door. On the coverage map, every pin carries its model’s optics, every cone is drawn at its reach level, and an entrance-gap check lists each entrance camera that falls short of identify in front of it, before anything is ordered. For an integrator, the entrance-gap list is a quote-ready list of cameras that fall short, drawn from the plan before any part is ordered. See it running against a 500-camera fleet on the interactive demo.