3

I work under Debian Linux and want to ortho-rectify and geo-reference images aquired by a tilted twin camera system using:

  • the attitude parameter (roll, ptich and heading) of the gyro-stabilized platform and

  • the position data (longitude, latitude, altitude)

from a parallel running IMU/GNSS logged at the mide-exposure time mark (see image below).

enter image description here

The flight platform is working over a sea surface, so terrain issues could be neglected. The transformation from longitude, latitude, altitude (ellipsiodal) to x, y, z (ortho-height GEOID) is done by the proj tool.

Normally I start to correct the perspective distortion ("ortho rectification") using openCV using

  • getPerspective to implement the passpoint configuration and get the rotation matrix setup and

  • warpPerspective to create the new image with a given resampling algorithm

and apply the geo-referencing aftermath

  • gdal_translate to write the ..gpc's and the SRS and

  • gdalwarp to create the GeoTIFF container with the expected resolution (resampling routine/ type again).

So the GDAL is here used as "GeoTiff-Driver" to implement the SRS metadata and to assure the resolution via passpoints which are calculated also in the getPerscpective process before.

To avoid two interpolation steps, based on the two tools and saving time and resources, I want do do this in one programatical step (tool) after reading the raw TIFF image into the RAM.

Question: Is either

  • the creation of a GeoTIFF (augmentation with the co-ordiante system metadata) using openCV or

  • on the other hand the perspective correction in GDAL (including the C-API)

in that way possible?

2
  • I'm not that experienced in orthorectification, but if my general understanding of your Q is approximately correct the least-time/resources method would be to apply all perspective and warp calculations to vrt first and then gdalwarp/translate at the from .vrt to geotiff. Nov 16, 2021 at 19:39
  • I am doing the same thing now and fine tuning with with QGIS. It seems to work well
    – Cary H
    Jul 27 at 17:26

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.