CADRG data in the polar regions does not display properly.

[gaopeng]

Their projection isinterpreted as GCS WGS84. It should be North Pole Stereographic.

[Even Rouault]

I think the fix is in trunk. See #2940

[warmerdam]

The relavent patch was r16750. Unfortunately it depends on being able to instantiate a coordinate transformer which is likely not possible in the ArcGIS environment (no PROJ.4). Consulting with Gao on how to proceed.

[Even Rouault]

Frank, in r18640 (trunk), I've fixed a leak of pszGCPProjection that was never freed

[warmerdam]

The transformation into AEQD projection was already in the 1.6-esri branch, however as noted it doesn't do anything in the context of ArcGIS since the transformer cannot be instantiated (nor PROJ.DLL).

As discussed with Gao, we will fallback to publishing the GCPPROJECTIONX metadata in the IMAGE_STRUCTURE domain similarly to what was done in the BSB driver. This essentially tells the projection the GCPs should be converted to in order to better represent the natural coordinate system of the image.

I have implemented this in trunk (r18637), 1.7 (r18638,r18643) and 1.6-esri (r18639, r18642).

In the process of working on this I also identified and corrected (#3347) which ultimately turned out not to be related.

[warmerdam]

Added a autotest of the polarsteregraphic case with and without PROJ (r18646) in trunk.

[gaopeng]

I am reopening this ticket with a new test case. It works for the north pole data. However, for the south pole the data shows up as WGS1984 rather than South_Pole_Azimuthal_Equidistant. Attached is a test image.

[warmerdam]

The code was supporting "zone 9" files which are north pole. I added "zone J" to the filter per the specification and now southern polar files seems to work properly too. The fix is in trunk (r18837), 1.7 (r18838) and 1.6-esri (r18839).