Skip to main content

Mitigate The Bowl Effect

The bowl effect (also called dome effect) is a distortion in aerial photogrammetry where the reconstructed model curves upward or downward at the edges.

Example of the bowl effect

Causes​

The bowl effect can be caused by multiple factors, but it's usually due to a suboptimal flight path or excessive lens distortion.

  • Nadir-only Images: capturing images pointing straight down
  • Straight line flight paths: such as when scanning roads or pipelines
  • Constant altitude: flying at a single altitude throughout the mission
  • Heavy lens distortion: from wide-angle or fisheye lenses
  • Rolling shutter: capturing images with a rolling shutter sensor while the drone is moving
  • Missing or corrupted GPS data: position data helps constrain the reconstruction. Without it, small errors compound into large distortions.
  • Featureless terrain: Uniform surfaces like water, sand, snow lack distinctive features for matching between images.

Prevention​

Use Ground Control Points​

By placing surveyed markers with known coordinates throughout the scene, you provide absolute reference points that anchor the reconstruction to reality. See getting ground control points.

Use RTK/PPK​

Drones equipped with RTK (Real-Time Kinematic) or PPK (Post-Processed Kinematic) positioning provide centimeter-accurate location data for each image. This high-precision positioning reduces the number of GCPs required.

Capture Oblique Imagery​

Include images captured at 85-80 degree angles in addition to nadir shots.

Fly at Different Altitudes​

Vary your flight altitude between passes.

Expand the Survey Area​

Capture imagery beyond your area of interest. Buffer zones around the edges help ensure accurate reconstruction in the target area, even if the edges exhibit some distortion.

Fly Cross-Grid Patterns​

Instead of flying parallel lines in a single direction, add perpendicular passes to create a cross-grid pattern.

Camera Calibration​

If you have processed another scene with the same sensor, you can transfer the camera calibration parameters from the good dataset to the bowl dataset, which in some cases can help. To do that, use both cameras and use-fixed-camera-params.