What is Pixel-to-Metric Calibration (mm per pixel)?

Pixel-to-Metric Calibration (mm per pixel)

Pixel-to-metric calibration is the step that gives an image a scale in millimetres per pixel. A length in pixels then becomes a length on the ground. Without it, an image shows that a feature exists but not how large it is.

How the scale is found

The simplest method uses a reference object. Divide its known length by its length in pixels. A 100 mm scale bar that spans 500 pixels gives 0.2 mm per pixel.

The camera geometry also gives the scale of a drone image. The ground sample distance (GSD) is the distance between pixel centres on the ground. For a camera that looks straight down, multiply the pixel size on the sensor by the height above the ground. Divide by the focal length. For example, a 4 micrometre pixel, a 24 mm lens and a height of 10 m give 1.67 mm per pixel. These values are only an illustration.

Tilted views

If the camera tilts, the ground distance per pixel grows and changes across the image. The general GSD formula uses the slant range to the ground point and the cosine of the slant angle. One mm per pixel value then holds only near one point. An orthomosaic gives one scale over the whole map.

Why it matters for pavement

Operators rate airfield pavement condition with the Pavement Condition Index (PCI), which follows ASTM D5340. The PAVER distress manuals set some severity levels by crack width. In asphalt, a non-filled longitudinal or transverse crack is low severity when its mean width is 6 mm or less. In concrete, a non-filled corner break crack is low severity when it is narrower than about 3 mm.

A pixel must therefore be much smaller than these widths for a width to be read from an image. At 1.67 mm per pixel, a 6 mm crack spans about 3.6 pixels.