[gdal-dev] How do GCP's in a VRT file work?

[Greg Troxel]

Joe Lovick via gdal-dev <gdal-dev at lists.osgeo.org> writes:

>        What is the recommended processing paradigm for slightly
> off-nadir imagery where their is a need for a full [3x3] matrix
> mapping to take it to a consistent projection? ie the images have at a
> minimum perspective effects.
>
>     the reason i ask, is i have a significant image pile that needs
> this re-projection to remove camera lens + perspective effects , to
> map it to a consistent conformal projection? Can i do this with
> gdal_warp? or do i need to code this up, and then go from their once
> their is just a rotation / scale left.

I have contemplated acquiring imagery and processing it, but so far only
contemplated.  My impression is that you need to resample, not just compute a
transform (hence gdalwarp vs vrt), because most things that use imagery
expect at most only transform, scale and rotation.

One thing that I found that may be useful is

  https://www.opendronemap.org/

however it seems to have "docker is the approach" which generally makes
me suspicious of build system hygiene.  That may just be spin and not
real, though.