Orthorectification is required to align imagery with a Map.

Normally this is done by the supplier of the imagery.

However imagery can be supplied without this process having been done.

In this case you have to do this yourself. There are quite a few Open Source tools that can perform this.

However you need the RCP data to do this or you need to create this information from the camera model.

This post is essentially to mark in time my thoughts so that I can retain IPR.

For UAV systems with cameras orthorectiification needs to occur in realtime to allow camera video or stills to overlay the mapping data correctly.

The simplest approaches are:

  • OpenGL using a "shadow mapping"/"image projection" approach to project the texture of each  frame onto a WGS84 ellipsoid with Terrain mesh. View looking straight down.
  • Using OpenCL/OpenGL to implement a reproection of the texture image.
While the above are not even 99% accurate the aim is to be close enough to achieve realtime display of the imagery on the map with a close enough image registration to the map.