Camera placement and pixel density

Whether a camera can identify a person depends less on its megapixel rating than on how many pixels land on that person. Pixel density, measured in pixels per metre across the scene, links the job a camera has to do with where it is mounted, how far it looks and which lens it uses.

What pixel density means

A camera's sensor has a fixed number of pixels across its width: 1920 for a typical 1080p camera, more for higher resolutions. Those pixels are spread across whatever width of scene the lens takes in. Point the camera at a doorway two metres wide and the pixels are packed tightly; point the same camera across a parking lot forty metres wide and each pixel covers twenty times as much ground. The calculation is simple:

Pixel density (pixels per metre) = horizontal resolution in pixels ÷ width of the scene in metres, measured where the target will be.

Because a lens's field of view widens with distance, pixel density falls the further the target is from the camera. A camera that identifies people three metres away may only show that someone is present at thirty metres. The useful question is never "how many megapixels" but "how many pixels per metre at the place this camera has to see".

Levels of detail

Security designers describe what a camera must achieve using a small set of levels, often abbreviated DORI: detect, observe, recognize and identify. The international video surveillance standard IEC 62676-4 gives pixel density figures for each level, and the figures below are the ones commonly used in Canada. Check the current edition and any requirement from your insurer or police service.

These are minimums for good conditions. Poor lighting, heavy compression, motion blur and rain all reduce the useful detail, so designers usually aim above the target for critical views. Reading licence plates has its own requirements, set mostly by plate size, angle and shutter speed.

Placing cameras for the job

Entrances and exits

Every person entering a building passes through a doorway, which makes it the best place to identify them. Narrow the view to the doorway itself, place the camera so faces are seen from the front rather than from directly above, and consider a second, wider camera that shows where people went next.

Height and angle

Cameras are often mounted high to protect them from tampering, but a camera mounted high and close to a door looks down on the tops of heads and the brims of caps. Pulling the camera back and lowering it, or using a longer lens from further away, keeps the vertical angle shallow so faces are visible.

Light and backlight

A camera inside a lobby looking out through glass doors on a sunny day sees bright sky and dark silhouettes. Wide dynamic range helps, but the better fix is placement: look across or into the space so the bright doorway is behind the camera or to one side. At night, a camera can only show what is lit, by site lighting or by its own infrared illuminator. Our guide to lighting for video surveillance covers this in detail.

Parking lots, yards and perimeters

Large open areas are usually covered for observation or detection, with identification saved for the choke points where people and vehicles must pass: gates, lot entrances and building doors. Trying to identify people across a whole parking lot takes far more cameras than most sites need.

A worked example

A lobby entrance has a pair of doors with a vestibule about 7.5 metres wide where people pass. A 1080p camera (1920 pixels across) set so its view spans exactly that width gives 1920 ÷ 7.5, or 256 pixels per metre, which meets a typical identification target. If the designer instead widened the lens to take in the whole 15-metre lobby, density would halve to 128 pixels per metre: good for recognizing known staff, not for identifying a stranger. The better answer is one camera tight on the doors and a second for the lobby.

The same camera across a 40-metre-wide parking lot gives 1920 ÷ 40, or 48 pixels per metre. That is enough to detect people and vehicles and roughly see what they are doing, but not enough to recognize anyone. If the owner wants to know which car entered, a separate camera at the lot entrance, set for plates, does the job better than a higher-resolution camera over the whole lot.

Checking a camera plan, step by step

Whether you are reviewing a contractor's proposal or your own design, the same checks catch most problems. Record the answer for each camera on the camera schedule so the plan can be checked again on site.

1. Write the camera's job. One line: identify people at the north door, observe the loading bay, detect movement along the east fence. Record the detail level it needs.
2. Mark the target point. Show on the plan where the person or vehicle will be when the camera needs to see them, such as the doorway or the gate lane, not the far wall.
3. Measure the scene width there. Use the lens's horizontal angle and the distance to the target point, or the camera maker's design tool, and record the width.
4. Calculate pixel density. Divide the horizontal resolution by the width and compare it with the target for that camera's job. Record both numbers.
5. Check the angle and light. Note the mounting height and the vertical angle to the target, any bright windows or doors in view, and how the scene is lit at night.
6. Confirm on site. After installation, stand at the target point, view the live image and an exported recording, and mark the camera as passed or note what needs to change.

Common mistakes

• Choosing cameras by megapixels without working out the scene width.
• Using one wide view to cover a whole room when identification is needed at one door.
• Mounting cameras high and close to entrances, so they see only the tops of heads.
• Pointing cameras at bright windows and doorways.
• Checking views on a sunny afternoon and never at night.
• Not writing down what each camera is for, so no one can check whether it works.

How ehZee can help

We start each camera layout by asking what each view needs to show, then check pixel density, angle and lighting for every camera on the plan and again on site after installation. See our video surveillance design service, and our guide to IP camera system design for the network and storage side.

Common questions