On 20. feb. 2017 21:03, juerg.maier wrote:

This link contains 2 versions of the land parts, extruded to center.
More precisely, the inner bottom has been shifted towards the center,
leaving the top as is.

earth_730_try2_extrude_1.stl is shifted 1mm down
earth_730_try2_extrude_5.stl is shifted 5mm down
If you ned

this was relative to a sphere radius of 99.479

https://www.expirebox.com/download/3ff25a2715fe8921704296de947a6b71.html

Now I wonder how you plan to print....

Carsten Arnholm

Hello Juerg,

I am confident that the result is not as good as we would like it to be,
but it may not be possible to do much better with the given data. I
don't know how accurate the original model is, and in any case
subtracting a sphere is just a zero order approximation sure to cause
the kind of anomalies observed. I live in Norway, so I iterated the size
and center of the sphere to make Europe look as reasonable as I could.
One side effect was that e.g. the gulf of Thailand became a
sub-continent. It does not seem possible to get it all correct with this
model, I can assure you I tried.

The determination of offset center was by trial and error, the argument
in favour of the offset is just that other values make things look much
worse. I also tried with different radius values. Again the result is
the best compromise I could find given the constraints. As mentioned
before, subtracting a sphere is wrong in any case. If you assume the
earth is one big ocean the shape of it is not a sphere, but an irregular
geoid. To improve on the coastlines you need to subtract the geoid, not
a sphere. Alternatively, make several different approximations
optimizing for different locations on earth and somehow find a way to
combine them. Not straight forward.

As for the "extrusion", I used my AngelCAD application which on the
surface is somewhat similar to OpenSCAD, but also different in important
ways. It uses AngelScript as the scripting language and Carve (
https://github.com/azarus/sybren-carve ) as the boolean engine instead
of CGAL. It allows importing polyhedrons from AMF files, so the first
thing I did was to use another piece of software I have and merge the
matching vertices of the original STL, then export as AMF. Then import
to AngelCAD and perform the subtraction of the sphere. Now I had the
"thin version" of the continents, before "extruding".

AngelScript is different from the OpenSCAD language. It is not just
declarative, it runs like most other languages and allows you to modify
variables. In AngelCAD, a polyhedron is an object that can be queried
and modified. A very useful feature is to be able to selectively modify
the coordinates of polyhedron vertices, you can effectively "morph" the
shape of an existing polyhedron. In this case I used the known radius
distance from the given sphere center to select which vertices to modify
and which ones to leave untouched, and simply moved the selected ones
towards the center. As you understand, this is not really extrusion
since the topology already is 3d and remains the unmodified in this
process. It is only the geometry (coordinates) that is modified. Once
complete, the model was exported to the STL you saw. I can show the code
if anyone is interested.

Not sure if this answers all your questions, but that is the story I
have.

As a side note, I am (very slowly) working towards the goal of releasing
my carve-based boolean engine to github. It is a separate console
application taking an XML file as input, does all the booleans and
writes STL, AMF etc. Obviously, it runs in a separate process from the
'host' application (AngelCAD in my case).

Carsten Arnholm

On 2017-02-22 08:53, juerg.maier wrote:

On 22. feb. 2017 18:51, juerg.maier wrote:

I didn't spend that much time, and it is an interesting problem. I will
check the earth_9_24.obj file (200MB), but one issue is that I don't
have reader code for the .obj format...yet. It can be converted with
meshlab though.

Howver, maybe you are better off with using only something like this
http://tf3dm.com/3d-model/world-globe-94593.html

That is a land-only model with well defined coast lines, no tiny islands
and a bit of thickness. Only problem is that the center offset from
origin again.... and the inside faces appear to be inverted.

See if you can use the model above first, meshlab will be able to
convert it STL. Perhaps you can difference it from a sphere in OpenSCAD.
If not I can have a look.

I don't quite understand how you plan to print. Are you printing both
STL's in one go with dual extruders? If yes, I think you just need
enough thickness of the land part to embed it in the ocean sphere?

If the topology is correct, it should work....

Carsten Arnholm

Don't forget that you could take a 2D vector path of the coastlines,
triangulate it in software of your choice, and transform into a sphere
using reverse projection of wherever the initial data was in
(Equirectangular, mercator, ...)

Then the earth oblate is not an issue.

Search for SVG World map. open in inkscape - export in format you can use.

On 2/23/2017 9:32 AM, Carsten Arnholm wrote: