RPCs in GeoTIFF

[rouault]

A proposal to embed RPC (Rational Polynomial Coefficient) is described at http://geotiff.maptools.org/rpc_prop.html
This has been implemented for years for example in the GDAL library.
Wondering if it might be of interest to be included in the GeoTIFF specification itself (it is rather orthogonal to the GeoTIFF encoding, using none of its tags and keys)

[kevcri]

Embedding RPC's definitely needs to be included in the GeoTIFF specification itself.
This is currently the case with NITF and therefore you can easily chip a section out of an image in NITF format and still retain accurate geopositioning. This also allows your software of choice to access proper sensor model information when doing an orthorectification.

Having RPCs embedded within the GeoTIFF specification will allow the entire community to realize the following benefits: -

1. Ability to create smaller image footprints from one larger image for training purposes while still having access to the sensor model info even if the chip crosses several "tiles".

2. Ability to use any software that supports reading and writing to GeoTIFF as opposed to NITF where in some cases writing out to NITF is a functionality left to a special "install" because software companies are reluctant to wait until the latest "NITF" testing to be done.

[EmDevys]

This is a proposal of interest for GeoTIFF and users of RPCs models.
However, this is an additionnal capability to the existing GeoTIFF, which needs some clarification / formalization (as well as example). One key question is what is the underlying GTModelTypeGeoKey? Should it be 0 ("unknown") or 2 (geographic 2D coordinate)?
I would suggest to elaborate such an extension as an extension or BP standard as soon as GeoTIFF 1.1 is done, if this requirement for extension is confirmed by the SWG, and one example is made available.