[OSM-talk] gdalwarp question

[Andy Allan]

On Wed, May 6, 2009 at 6:44 PM, Torsten Mohr <tmohr at s.netic.de> wrote:
> Hello Jukka,
>
> no, thanks for your help, any hint and discussion is really appreciated.
>>
>> Sorry, I misunderstood a bit what you were going to do.  It may well be
>> that for Mapnik you'll need to reproject raster image first. I do not much
>> about Mapnik and while a have been using gdal utilities very much I don't
>> believe I have ever needed to create output in Google projection.
>> By first look what you have done does make sense. You can make a couple of
>> further tests:
>
> I solved the issue now in a different way.  Based on information on wikipedia
> about mercaator projection i did the reprojection of the blue marble myself.
> It took quite a while to execute, but the reprojection itself worked quite
> fine.
>
> Now the borders (shoreline_300) are exactly on the borders of the continents
> and islands on the "blue marble".
>
> The relevant part of the script i use is attached below, if anybody is
> interested please write me an email (i use an own library to handle the
> images themselves, that part needs to be adapted to e.g. PyImage).
>
> Thank you all for providing me help, i hope the script below is of value to
> anybody.

As a point of comparison, I did the following last weekend
(coincidently whilst you were battling with the same issues).

1) Download GeoTiff images from
http://www.unearthedoutdoors.net/global_data/true_marble/download
2) Remember from bitter experience that mapnik doesn't do on-the-fly
reprojection
3) Gdalwarp them into spherical mercator.
4) Remember from bitter experience that gdalwarp messes up the geotiff
coordinates (i.e. they come out as non-spherical mercator and then
confuse the gdal plugin for mapnik)
5) Calculate the extents of the image using gdalinfo on the orginals
(using mapnik prj.Forward code in a simple python script)
6) use mapniks' raster plugin with lox/loy/hix/hiy set using the output from 5
7) ... fiddle around a bit with layers and coastlines ...
8) Admire imagery and OSM combined at
http://www.opencyclemap.org/?zoom=4&lat=42.32525&lon=3.33555&layers=B000

:-)

Cheers,
Andy