More Coordinate Issues

[dtmicro]

I've tried Gdalwarp and extracting coordinates directly out of the KML files from the The National Map downloader's KML files, I get the same results with both, coordinates that are off by about 10-20 feet in some spots. Some roads do not line up exactly on connecting tiles.

Anyone have any idea as to what the deal is, do I need to buy Global Mapper?

 

[arno]

Hi,

What are you comparing to when seeing the offset?

Did you use the right projection when processing the data? I know sometimes even the wrong unit can give an offset (like feet vs. survey feet).

 

[dtmicro]

Hey Arno,

Thanks for asking, I can tell by the roads and airports and what not. It's actually off by more than 20 feet, hard to tell since it's not off in just one direction, it appears to be off in both X-Y. I also double checked it against a payware scenery in the same area, so I confirmed the error for certain.

To answer your other question, yah I've been trying all kinds of things.

Here is what I've tried:
1) Custom Generating Xdim, Ydim in my .NET code instead of using Gdalinfo's XDIM/YDIM
Hence, xdim = (W-E) / (imgWidth -1), and yDim = (N-S) / (imgHeight - 1)

2) Use the KML / KMZ file provided by NAIP at the end of the download email instead of Gdalinfo after a GdalWarp.

3) Tried 10+ variations of Gdalwarp (running one more now). This is on Nad83 files, haven't tried converting the Mercatum_Spherical ones to WGS84, as that is a whole different thing, just trying to get these coords to match up perfectly. I do not recall having this issue with the older Geo-Tiff format.

Here is my latest concoction:

"C:\Shared\GdalWarp\gdalwarp.exe" -t_srs "+proj=latlong +datum=WGS84" -r cubic --config GDAL_CACHEMAX 2000 -wm 2000 -co INTERLEAVE=PIXEL -co TFW=YES -of GTiff -rcs "m_4011102_ne_12_1_20110810_20111011.jp2" "BearLakeExt_Tile1.tif"

 

[dtmicro]

I guess the only things left to try are:

I read somewhere that someone said if you don't inject the coordinates directly back into the .TIF to re-assign it as a Geo-Tiff (rather than just directly entering the coords w/ batch), that the results can be inconsistent. So maybe that is why, I guess I'll try that next.

Otherwise, it's gotta be something with my conversion and projection I guess. Either that or these JP2 files have bogus coordinates due to rounding precision when they converted them(but I doubt it, just throwing it out there).

 

[arno]

Hi,

I usually just work with GeoTIFF files, so that the INF file is very easy. The gdalwarp you use seems OK, unless the projection meta data of the source image is not accurate of course.

 

[dtmicro]

Is there a place to still download the Geo-Tiffs from NAIP imagery?

Well, I finally got it working, and you're going to laugh what the problem was... Apparently I had a MIXED version of a VERY OLD GdalTranslate and semi-newer Gdalwarp (different versions), and it appears after deleting and reinstalling a newer version of both, the problem went away somehow.

I wasted too much time on that, totally sucked. I didn't realize I mixed the Gdalwarp versions. Another issue was the larger JP2's are Web Mercator Aux Sphere format, but only 20% of them are in that format, so I just put those aside for later. It was just a combo of hellish carelessness that caused it. When copying all the BGL's back and forth for testing, I accidentally merged some incompatible directories. Now I just want to fly myself into the ground in a B-52.

 

[dtmicro]

Just FYI:

It turned out NOT to be my re-installation of Gdalwarp that was the issue (was a coincidence), the issue is something like compiling tiles in different sets (such as if compiling one tile separately from another), seems to produce slightly varying coordinate offsets (very odd). Will have to experiment more to see what's going on.