Image Quality Requirements for an Effective Crack Detection System

Considerations for an effective image based Crack Detection System

For effective crack detection, specific image features are required to allow effective automatic or manual crack assessment. Without these features, the performance, repeatability and quality of the crack mark-ups produced will be affected.

Requirements for an effective image based Crack Detection System

  1. A high resolution camera image: The minimum crack width detectable by a system is related to the resolution of a camera system used to detect the cracks. In specific conditions cracks can however be detected that are lower than the resolution of camera.
  2. An image unaffected by ambient lighting conditions: When only sunlight is used, the image quality can be severely affected by the ambient lighting conditions. In some conditions the results are perfect, with clear shadows formed on a surface. In other conditions imagery is less effective. The primary problems include:
    • a. Shadows from trees, builds and roadside assets impinge onto the surface image. Most standard asset cameras do not have the dynamic range to cope with this type of conditions. Thus cracks are not detected effectively in the shadowed regions.
    • b. Lower lighting conditions. Effective crack detection relies on a deep shadow being formed. On overcast days, or later in the day, these smaller cracks are often not detected.
  3. A shadow: Just having artificial illumination of a surface with a high resolution camera is not sufficient to produce an image where the cracks are detectable. For a crack to be seen, it needs to develop a shadow for those cracks. This is because a crack is only observable by the contrast of the darker shadow regions from the surrounding road texture. A camera will tend to flatten an image, thus if the lights are shining down a crack, instead of across a crack, then the crack will essentially disappear within the image. The direction of the lights therefor defines the crack orientation that the system is more sensitive to.
  4. Full Lane Width Scanning: To obtain a useful map of the road surface, a system needs to survey the entire surface. Cracking does not always just occur in the wheel paths. For example distress cracking often form between lanes, or starts from the edge of the road. As a result only surveying part of a lane is not necessarily representative of the entire lane.

It should be noted that the detectability of a crack on a road surface depends on the actual road surface. This is because a crack is detected by the shadow a crack makes in contrast to the different shades of the actual surface material. This is true for both manual and automated means. As a result, if the surface is relatively consistent, then a crack will be easy to detect, however if a surface has high contrasts then cracks are harder to detect.

For example with a concrete surface, cracks significantly lower than a systems resolution can be detected (we have found cracks as small as 1/10mm). For asphalt, cracks are also easier to detect, due to the more consistent dark grey surface colour (for asphalt we find that the 50% detection limit is around 0.5mm). Spray seal roads, due to the often high brightness of the rock in contrast to the dark bitumen, are the hardest to detect. As a result it is difficult to detect cracks smaller than the resolution of the camera for spray seal roads (e.g. for our system 50% detection limit is around 1mm).

Crack detection example 1

Crack detection example 2

Crack detection example 3

Return to top