Orientation quality & Aerotriangulation


Orientation parameters

The orientation parameters or Aerotriangulation result consists of camera position, rotation and inner camera parameters (focal length, principal point, distortions). Next to the images, it serves as input for SURE and is defining the output location of the data and with its quality the output quality.

Summary

  • the quality of the orientation has a strong impact on the quality of the final output of SURE
  • we recommend ensuring a one-pixel-parallax or better orientation
    • Aerotriangulation, Structure from Motion or Bundle Adjustment software output needed 
    • choose an option in the orientation solution to have tie points used in every image
    • orientation data retrieved from GPS/INS alone (no tie points) is insufficient
    • also for oblique images one-pixel-parallax should be achieved
    • most simple tool for QA is a stereo viewer for checking the shift in vertical direction
  • your benefits
    • positional accuracy is higher
    • better edges
    • more detail
    • less surface noise

Further remarks

  • one-pixel-accurate orientation is relevant for all images
  • SURE works also with less good orientation, but results will be affected negatively
  • a weak orientation will manifest in surface noise, missing detail and unsharp edges
  • one-pixel-accurate orientation can only be achieved, if tie points are used on every image
    • solutions, where tie points are only used and matched for some images should be avoided
      • e.g. avoid using relative orientation of camera rig only (tie points in Nadir only etc.)
      • many solutions offer multiple strategies - please consult your AT software vendor
    • quality check of the AT is recommended by examplary stereo measurements in the block
      • measure the remaining parallax perpendicular to epipolar direction
      • the sigma-naught value (a priori sigma or error of image observations) can be misleading
        • images might not have enough tie points between each other value might still be good
        • solution: double check also the number of tie points between of Nadir to Oblique images
        • apply stereo measurement method also between images of different view direction
    • additional runtime will lead to higher accuracy, more detail, better edges and less noise

Accuracy 

  • using automatic tie points / feature points improve the relative accuracy between the camera
  • the final positioning accuracy of your data is directly depending on the position and quality of Aerotriangulation
    • for optimal absolute accuracy of position, ground control points should be used during Aerotriangulation
    • ground control points should also be used when camera self calibration is applied 
      • e.g. for classic close range and UAVs, where the camera parameters are estimated in the project
      • for pre-calibrated metric cameras with stable sub-pixel accurate interior orientation
        • e.g. large frame metric aerial cameras or stable pre-calibrated metric cameras
        • results are generally better also in presence with less ground control points
        • ground control points should still be used in good distribution for good positioning
      • for QA, check points (ground control points not used in adjustment) should be used
    • ground control point accuracy should be at least 4x higher than the desired final accuracy