DORI means Detection, Observation, Recognition and Identification. It helps you plan how much detail a security camera should capture at a chosen location. A DORI distance is a pixel-density guideline, not a guarantee of a recognizable face, readable plate or successful AI detection.
- DORI describes four visual-detail tasks in the older IEC 62676-4:2014 framework; newer specifications may use the 2025 framework instead.
- Pixel density is expressed in pixels per meter (px/m) or per foot (px/ft), rather than a percentage accuracy score.
- Start with the detail needed at the target, then consider its distance, the width of the scene, lighting, focus and viewing angle.
- External lighting can enhance footage, especially in low-light scenarios.
A 2MP camera can provide useful detail across a narrow entrance while the same resolution spread across a large parking lot may be insufficient for faces. Decide which area needs identification and which only needs an overview before choosing a lens or resolution.
How To Use DORI When Selecting Security Cameras
The familiar DORI planning figures are 25, about 63, 125 and 250 pixels per meter for detection, observation, recognition and identification respectively. Check which standard a datasheet uses, and allow a margin for real conditions rather than treating a threshold as a pass/fail accuracy result.
When selecting security cameras for your home or workplace, it’s important to set the stage. This involves figuring out where you will be installing the camera and the amount of information you are hoping to glean from it.
Here’s how:
- First, state the task: do you need to notice someone, describe their activity, recognize a person seen before or identify them? Treat automatic facial recognition and number-plate reading as separate application requirements.
- Measure the horizontal width you need to cover at the target’s position, in meters or feet. This scene width is different from the camera’s angular field of view in degrees, and it increases with distance.
- Finally, calculate the horizontal recording resolution needed for the chosen pixel density and scene width. You can also start with the stream resolution and work out the maximum scene width, as the pixel-density example below shows.
How to Use DORI When Designing a CCTV/Security Camera System
DORI helps divide a site into different jobs. Use an overview camera for context and a suitably framed view for identification where needed; one wide image may not provide both at every distance.
Although this is a spec that is related to security cameras entirely, there are some other things we can do to improve the footage of the entire system starting with improving the scene lighting.
Depending on the range you are specifying for accuracy, you can opt to set up external lighting sources to help improve the camera footage, especially at night. Depending on which low-light system your camera will be using the most (either Infrared or full-color night vision) it will be best to install external IR illuminators or visible white light sources to help the cameras capture better footage.
For a parking lot, a fixed overview can show where people and vehicles move, with another camera framing the entrance or payment point more tightly. A PTZ can inspect selected areas, but it only supplies that narrow detail where it is aimed at the time. Do not assume a moving PTZ replaces continuous coverage of every entrance.
Detection
Detection is the lowest of the four traditional DORI levels: 25 px/m, or about 8 px/ft after conversion. Its purpose is to establish whether an individual is present.
At this level, plan to notice presence rather than identify a face. A distant figure may occupy too few pixels for reliable clothing or facial detail, especially in difficult lighting.
Do not use a detection-distance figure as the operating range of a person-detection algorithm. Analytics have their own size, angle, contrast and scene requirements. Processing may make a picture easier to view, but it cannot guarantee that missing identifying detail will be recovered.
Observation
Observation uses about 63 px/m, approximately 19 px/ft by conversion, often rounded to 20 px/ft in practice. It is intended for seeing activity and characteristic details rather than identifying a particular person.
An observation view can help you follow an activity, such as someone entering a yard or approaching a door. Distinctive clothing may help describe what happened, but visibility still depends on lighting, contrast and movement.
Compared with a presence-only view, you are planning for more context: how many people are visible, their activity and useful clothing details. That can be enough for an overview camera even when another camera provides a closer identification view.
Observation should not be sold as guaranteed face recognition or plate reading. If either is the purpose of the installation, set a specific requirement and test it with recorded examples at the intended position.
Recognition
Recognition uses 125 px/m, about 38 px/ft by conversion, commonly rounded to 40 px/ft. The task is to judge whether the person shown is someone the viewer has seen before.
A familiar person may be recognizable from a combination of visible features, but pose, darkness and motion can still prevent a useful result. Check the saved stream at the quality and view size the operator will actually use.
Number-plate capture needs its own planning. Plate angle, character size, headlights, exposure and vehicle speed can defeat a scene that otherwise meets a DORI figure. Nor does a human-recognition threshold certify an automated face or plate-recognition system.
Identification
Identification uses 250 px/m in the traditional DORI model, about 76 px/ft by conversion and often rounded to 80 px/ft. It is a planning guideline for identifying an individual, with real-world margins still needed.
There is no universal minimum distance of 20 meters or 65 feet for identification. The distance depends on the actual image width, lens and scene geometry. A close entrance view can meet the planning density at a much shorter distance; a very wide view can miss it even with a high-resolution camera.
What is Pixel Density/PPM/PPF?
Pixel density describes how many horizontal image pixels represent a meter or foot across the scene at the target’s location. Use the resolution of the stream you record or review, not just the sensor’s advertised megapixel count.
More pixels across the target can preserve more detail, provided focus, exposure, lighting and compression are also adequate. Digital enlargement alone does not increase the detail originally captured.
For a simple estimate, divide horizontal image pixels by the scene width at the target, using the same distance unit throughout. A narrower view or a higher horizontal resolution increases that estimate. A lens calculator or pixel counter can help check it, especially where lens distortion changes density across the frame.
A Pixel-Density Calculation You Can Check
Suppose a recorded stream is 2,560 pixels wide and covers 16 meters across the entrance at the target’s position. The estimated density is 2,560 ÷ 16 = 160 px/m. That is above the traditional recognition figure of 125 px/m, but below the identification figure of 250 px/m. This is an illustrative calculation, not a test of a particular camera.
For the same 2,560-pixel stream to reach 250 px/m, the scene width would need to be about 10.24 meters or less at that position: 2,560 ÷ 250 = 10.24. Narrowing the lens view may achieve that, but it leaves less of the entrance in frame. If you need both a wide overview and identifying detail, use separate views rather than digitally enlarging the overview.
Check the furthest important point and the edges as well as the center. Test a person walking through after dark, replay the saved main stream and inspect the face at a usable display size. Lighting direction, shutter blur and aggressive compression can spoil detail even when the calculation looks sufficient.
Why Newer Datasheets May Use Different Terms
IEC 62676-4:2025 introduces a different set of visual tasks, ranging from overview to scrutinize. For example, its characterize category uses 250 px/m, while validate uses 500 px/m. These are not interchangeable labels for the four older DORI tasks. Confirm the edition and task required by the project before comparing quoted distances.
The newer framework also distinguishes human interpretation from software analytics. Keep a manufacturer’s DORI range, its AI detection range and its IR illumination reach separate: each answers a different question. For an existing system, record which framework you are maintaining so a later camera replacement is compared on the same basis.
Conclusion
Use DORI to state the job each view must do, calculate a starting pixel density and check the actual recording. Keep an overview where context matters and reserve a tighter view where identification matters. A paper specification is useful for planning, but a moving-subject test in the real lighting decides whether the installation provides the evidence you need.