Colin Marquardt pisze:
2009/3/21 Dane Springmeyer <[email protected]>:
These patches are not yet in svn.
They are now since Artem committed them in
http://trac.mapnik.org/changeset/1027, so they will be in Mapnik
0.6.0, great.
Great :)
Marcin, would you mind describing a bit more which exact parameters
you used for your nice Poland maps?
In osm.xml I use files from attachments: dem_styles.xml and
dem_layers.xml. There are also defined scaled versions to avoid time and
memory consuming downscaling.
Layer dem* are just after coastlines, "leisure-big" and "leisure" (all
forests, industrials, aeroways and any areas other then water). I have
almost all polygons without transparency, only housing and city range
areas are transparent:
http://mapa.ump.waw.pl/ump-www/?zoom=12&lat=52.99057&lon=18.49656&layers=B00000T
So basically You draw "background" layers that should be shaded, then
shading layer and then the rest (roads, icons, lakes, rivers, other
polylines ...).
Hill shading is done using new merge method: grain_merge, where output
pixel value are evaluated using:
I+M-0.5
where I is intensity of source (already drawn layers) and M is mask,
both in 0..1 scale.
I would suggest experimenting also with multiply2, because with some
gimp corrections (increasing or decreasing contrast/gamma in <0.5 and
>0.5 range), could give better effects (less influence to color, only
intensity shading).
Both grain_merge and multiply2 makes background unchanged for shading
mask value: 0.5 (or 128 in 0..255 scale).
As for preparations:
gdal_merge.py -v -o ../srtm.tif *.hgt
gdalwarp -of GTiff -srcnodata 32767 -dstnodata 32767 -s_srs \
"+proj=latlong +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 \
+x_0=0.0 +y_0=0 +k=1.0 +units=m +nadgri...@null +no_defs +over" \
-t_srs "+proj=merc +a=6378137 +b=6378137 +lat_ts=0.0 +lon_0=0.0 \
+x_0=0.0 +y_0=0 +k=1.0 +units=m +nadgri...@null +no_defs +over" \
-rcs -order 3 -tr 76.437 76.437 -wt Float32 -ot Float32 \
srtm.tif warped.tif
./hillshade warped.tif hill_8.tif -z 3
Adjust -z to Your needs and land characteristic.
Latest gdal is able to correctly interpret hgt files, just naming is
important.
I don't remember what I used to fill holes in hgt files, but it was
probably some Windows or commandline application. I did all of
preparations once, and now just use generated tiffs.
hillshade is basically from demtools with simple modification to move
mean value (ie. gray value of flat land) to middle of scale (128) by
adding correction:
cang -= sin(alt*degreesToRadians)-0.5;
This way grain merge will leave intensity unchanged on flat areas and
mean intensity of map will be similar to one before shading.
Here is version I modified and used:
http://marcom.homeip.net/ump-www/demtools.tgz
You could also skip this fix and use multiply merge method. Mean output
intensity will be lower, but without areas overflowing intensity scale:
http://mapa.ump.waw.pl/ump-www/?zoom=13&lat=49.79812&lon=19.80539&layers=B00000T
Choose what You find best for Your map.
To generate lower resolutions images You could probably use different
values for "-tr 76.437 76.437" gdalwarp command (power of 2), ie:
152.874, 305.748, ..., but scaling parameter to hillshade command is
needed accordingly. I haven't tried it anyway, just scaled tiff using
gimp and forced proper geotags with gdal :)
I hope this is enough, because I didn't understand some of the terms You
used in Your questions, so I only described what I did :)
For areas bigger then Poland (are there bigger Countries then Poland? ;)
) problem could be aligning separate rasters. I use one big tiff (200MB)
and gdal should allow much bigger files, but I'm not sure if mapnik
does. I didn't try using adjacent rasters for hillshading, but expect
artifacts (~1px strips with double hillshading effect?). Probably more
careful aligning to image resolution would be needed: making sure pixels
from rasters are aligned with pixels of output. Now I only matched
resolutions (magic 76.437 value), to avoid artifacts when upscaling.
BTW. as for big tiffs, I have some problems with segmentation faults. I
believe its thread unsafe libtiff (60% of stacktraces are there
originated) on mapnik initialization step, when big geotiff is analyzed
for the first time (reading tiff headers and tags?). The bigger tiffs,
more time it takes to initialize and bigger chance for segmentation
faults. If renderd finally reads osm.xml file in each thread, everything
is fine, only parsing/loading step(load_map(...) ) is multithread
unstable. When I added 2-3s delay between thread creations, I have
almost no such problems, so I ignored it meantime.
I think with latest(1 month ago svn) mapnik and mod_tile/renderd I had
the same problem & "solution" (delayed thread creation). Anyway for day
to day generating Poland I still use self-patched mapnik-0.5.1, and
quite old mod_tile apache module.
Regards,
Marcin Rudowski
<Layer name="dem30" status="on" >
<StyleName>raster30</StyleName>
<Datasource>
<Parameter name="type">gdal</Parameter>
<Parameter name="file">&geotiffs;/hill_30n.tif</Parameter>
<Parameter name="format">tiff</Parameter>
</Datasource>
</Layer>
<Layer name="dem15" status="on" >
<StyleName>raster15</StyleName>
<Datasource>
<Parameter name="type">gdal</Parameter>
<Parameter name="file">&geotiffs;/hill_15n.tif</Parameter>
<Parameter name="format">tiff</Parameter>
</Datasource>
</Layer>
<Layer name="dem8" status="on" >
<StyleName>raster8</StyleName>
<StyleName>raster4</StyleName>
<Datasource>
<Parameter name="type">gdal</Parameter>
<Parameter name="file">&geotiffs;/hill_8n.tif</Parameter>
<Parameter name="format">tiff</Parameter>
</Datasource>
</Layer>
<Style name="raster4">
<Rule>
<MaxScaleDenominator>50000</MaxScaleDenominator>
<RasterSymbolizer>
<CssParameter name="opacity">1.0</CssParameter>
<CssParameter name="scaling">bilinear</CssParameter>
<CssParameter name="mode">grain_merge</CssParameter>
</RasterSymbolizer>
</Rule>
</Style>
<Style name="raster8">
<Rule>
<MaxScaleDenominator>250000</MaxScaleDenominator>
<MinScaleDenominator>50000</MinScaleDenominator>
<RasterSymbolizer>
<CssParameter name="opacity">1.0</CssParameter>
<CssParameter name="scaling">bilinear</CssParameter>
<CssParameter name="mode">grain_merge</CssParameter>
</RasterSymbolizer>
</Rule>
</Style>
<Style name="raster15">
<Rule>
<MaxScaleDenominator>1000000</MaxScaleDenominator>
<MinScaleDenominator>250000</MinScaleDenominator>
<RasterSymbolizer>
<CssParameter name="opacity">1.0</CssParameter>
<CssParameter name="scaling">bilinear</CssParameter>
<CssParameter name="mode">grain_merge</CssParameter>
</RasterSymbolizer>
</Rule>
</Style>
<Style name="raster30">
<Rule>
<MaxScaleDenominator>5000000</MaxScaleDenominator>
<MinScaleDenominator>1000000</MinScaleDenominator>
<RasterSymbolizer>
<CssParameter name="opacity">1.0</CssParameter>
<CssParameter name="scaling">bilinear</CssParameter>
<CssParameter name="mode">grain_merge</CssParameter>
</RasterSymbolizer>
</Rule>
</Style>
_______________________________________________
Mapnik-users mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/mapnik-users
