NITF Improvements Offered

[warmerdam]

Frank,

I have just begun using GDAL to work with NITF images. Nice package. Thankyou. I needed a few new features and added them. Attached is a patch file you can do with as you will. The changes do this:

 $ Add NITF_ICORDS to the dataset's metadata.
 $ Add all ICHIPB fields to the metadata.
 $ Read RPC00A exactly as RPC00B is read.

(I've never been able to see any difference between the A and B versions. The gov't's

STDI-0002-version-2.1-16-Nov-2000.pdf document says they differ in some coefficient ordering, but comparing with STDI-0002-version-1.0-25-Aug-1998.pdf, I see no difference at all.)

I think you might want to move the ICORDS change to a different spot and change the comments. There are also /*DRM:*/ comments you'll want to remove. They mark every change.

FYI, I needed NITF_ICORDS to interface with some annotation feature of IMAGINE with a NITF reader module.

Here are some possible bug/deficiencies I noticed while poking around:

• The NITF_MIN/MAX_LAT/LONG metadata items look wrong. The RPC region usually straddles LAT/LONG_OFFSET by LAT/LONG_SCALE. However, the metadata items begin at LAT/LONG_OFFSET and extend from there by 2*LAT/LONG_OFFSET
• The RPC projection code doesn't know about ICHIPB and so will incorrectly project points for a NITF file with ICHIPB.
• The ERR_BIAS and ERR_RAND RPC00B fields aren't put in the dataset's metadata.

Thanks again for GDAL,

Doug

[warmerdam]

Provided patch

[warmerdam]

This needs to be reviewed and considered for 1.5.0. NITF is important to our sponsors.

[warmerdam]

Frank,

One more thing, my email didn't disappear, I just couldn't find it right off. The source of the coefficient ordering was from the files ossimRpcModel.cpp and ossimNitfRpcModel.cpp from ​http://www.remotesensing.org/ossim/

Searching the code for thePolyType shows the interesting bits.

Doug

[warmerdam]

Deferring to 1.6. Also of note may be #2040 about RPC00A vs. RPC00B.

[Even Rouault]

I think that all what was contained in the patch has been implemented. ICHIPB reading was added in r6864 (in a slightly different but equivalent form from the patch)

[hobu]

Even, this is a 1.7 ticket now, correct?

[Even Rouault]

To me it looks like this ticket could be closed as AFAIK all elements in the patch have ended up being incorporated in gdal, but I'll let FrankW confirm. But for sure, this is certainly not a 1.6.1 ticket.

[warmerdam]

I believe this to be the case. Closing.

I feel bad we didn't make better use of Doug's contribution.