Megapixels describe the image, not the evidence at your gate. Pixel density connects resolution to the width of the real scene, letting you decide whether a view is for awareness, recognition or identification before cable is installed.
Start with the evidence task
Begin with an operational requirement written in plain language: detect a person crossing a boundary, recognize a known employee, identify an unknown face, or read a vehicle plate. Those are different jobs and often need different cameras. Mark the exact line where the subject must be useful and measure the distance from the proposed mounting point. Then define the maximum scene width at that line. A camera watching a 20-metre-wide yard may provide excellent awareness yet too little facial detail for identification. If both overview and identification matter, use two views. Asking one ultra-wide lens to do both usually produces a beautiful live screen and disappointing evidence.
Calculate pixels per metre
Pixel density is horizontal image pixels divided by scene width. A 3840-pixel image covering 15 metres delivers about 256 pixels per metre at that plane; the same image stretched across 30 metres delivers about 128 px/m. The calculation is simple, but only valid at the measured distance and for the configured stream resolution. Axis’s interpretation of IEC 62676-4 uses approximate DORI planning levels of 25 px/m for detection, 63 for observation, 125 for recognition and 250 for identification. These are planning thresholds, not guarantees. Angle, compression, motion, lighting, focus and subject position can still make evidence fail. Treat the number as a design gate, then validate the result in the installed scene.
Choose the lens with the exact sensor
Lens choice controls how those pixels are distributed. A shorter focal length shows more width and makes every subject smaller; a longer focal length narrows the view and concentrates detail. Sensor format also affects field of view, so focal length alone cannot predict the image across different models. Use the manufacturer’s lens calculator or field-of-view table for the exact camera and sensor, then confirm on site. Varifocal cameras make commissioning easier because the lens can be adjusted to the measured zone, but do not use the zoom ring to compensate for a poor mounting position. Leave enough context to show how a person reached the target while keeping the required detail at the evidence line.
Control angle and perspective
Perspective matters as much as width. A steep camera angle hides faces under caps and changes the apparent dimensions of objects. Mounting very high can protect equipment but may defeat identification. At doors and gates, position the dedicated detail view close to expected head height and as near to frontal as practical, while a separate overview camera provides continuity. For vehicle plates, control both horizontal and vertical angles and reserve a tight lane view; a general driveway camera rarely handles fast motion, headlights and plate exposure at once. Draw the viewing cone on a plan, mark near and far evidence lines, and note where columns, signs, foliage or open doors can block the subject.
Calculate from the recorded stream
Calculate with the recording stream, not the marketing maximum. Some recorders accept a lower resolution, transcode remote video or use a substream for live grids. Digital zoom cannot create detail that was never recorded. Compression also matters: heavy H.265 quantization can preserve a clean background while smearing a moving face, especially in rain or low light. Confirm focus across the required depth, disable any setting that silently changes resolution, and inspect exported frames from the recorder—not only the camera’s direct browser view. If analytics are part of the goal, verify their separate minimum object-size and pixel requirements because they may differ from human identification criteria.
Validate with moving subjects
Commission in the worst credible conditions. Place a real person at the near and far limits, have them walk across and toward the camera, and export clips in daylight and darkness. Use a printed face chart or measured target only as a repeatable reference; the decisive test is whether the required task can be performed from recorded moving video. Check backlight, rain, vehicle headlights and the camera’s night shutter. Save annotated reference frames with the subject distance, lens position, stream resolution and exposure settings. Those records prevent a later ‘tidy’ re-aim from quietly removing the evidence zone.
Create a coverage schedule
Use a coverage schedule for handover: camera name, purpose, target line, distance, scene width, recorded horizontal resolution, calculated px/m, day result, night result and known limitations. Recheck it after landscaping, shelving, signs or lighting changes. The most economical design is not the camera with the fewest units; it is the design that uses overview cameras for context and narrower cameras only where detail has a defined value. Measurable coverage makes that trade-off visible before purchase and defensible after an incident.
Work a complete coverage example
Work a sample before approving the whole layout. Suppose an entrance must provide identification across a 4-metre-wide capture zone and the selected camera records 2688 horizontal pixels. The nominal density is 672 px/m, comfortably above the planning threshold, but that number applies only at the plane where width was measured. If the view expands to 12 metres at the far pavement, density there drops to 224 px/m before perspective and image-quality losses. The practical design may therefore use the near zone for identification and label the far zone as recognition only. Put both boundaries on the drawing and repeat the calculation whenever lens, resolution or mounting point changes. This stops a single optimistic ‘maximum distance’ from being copied into a quote, risk assessment and handover pack as if it described the whole scene.
Example: 3840 horizontal pixels across 15 metres equals 256 px/m. Across 30 metres it falls to 128 px/m. Doubling scene width halves density even though the camera is still labelled 4K.
Before you sign off
- Write detection, observation, recognition or identification goal
- Measure target distance and scene width
- Calculate pixels per metre from recorded resolution
- Use the exact camera sensor and lens data
- Control camera angle and obstructions
- Export moving-subject tests by day and night
- Save an annotated coverage schedule
Can another person prove the system still works?
Record the final view, night image, bitrate, alerts, firmware and access method. A system is not finished until the owner can verify recording and export without the installer standing beside it.
Every site is different. Confirm manufacturer instructions, electrical requirements, privacy obligations and local regulations before installation.
Sources used for this guide
Primary standards and manufacturer documentation are linked so you can verify thresholds, features and current product behaviour.
Frequently asked questions
How do you calculate CCTV pixel density?
Divide the image’s horizontal pixel count by the real scene width at the target plane. For example, 3840 pixels across 15 metres is about 256 pixels per metre.
How many pixels per metre are needed to identify a person?
A commonly used IEC 62676-4 planning level is about 250 px/m for identification, but angle, light, movement, focus and compression still require an installed-scene test.
Does a 4K camera identify faces farther away?
Only when the lens and scene width put enough recorded pixels on the face. A 4K ultra-wide view can have less useful subject detail than a narrower lower-resolution view.
What is DORI in CCTV?
DORI groups pixel-density planning into detection, observation, recognition and identification. It helps translate a coverage goal into a measurable design but is not a guarantee of evidential quality.