Real time 3D map using WebGL

marek kleciak · 2013-10-30 16:36:47

No problem. I believe You can fix it in near future. Ther are some important point of Interest whoch need it (stadions).

regards,
Marek

things-change · 2013-10-31 08:20:14

j3d wrote:

@things-change: Not sure how to name this kind of roof in natural language. The problem is more the algorithmic part. The first question is how to subdivide the outline into rectangles. Method 7. described in http://wscg.zcu.cz/wscg2003/papers_2003/g67.pdf could produce the result you wish but it also requires to provide the exact outline and height
...

Maybe there is no name for this kind of roofs. As descriped in the document in Point 7, these forms are grown up through the addition of extensions f.e.

The method ist very interesting an allows a lot of shapes to create. But it could be to specific for the OSM commuity. But I think even this kind of modelling has no possibility to create the roof that I have posted. Until now I've seen nowhere a 3D model for this roof.
The point is that there is a ridge which is not at the top of roof but splitted from the middle of one side. I fear even even the descriped process in the document would create a splitted roof.

j3d · 2013-10-31 18:31:45

Marek, with S3DB there are still lots things that need to be addressed on the implementation side before I would approach more advanced themes. This could be worked out on the wiki in my opinion. If there will be a meeting I might attend if time permits, but for future 3D development I'm staying a side-line viewer for now

j3d · 2013-10-31 18:32:57

... maybe should start a 3D mailing-list instead of being constantly off-topic in this thread

cmif4 · 2013-10-31 18:39:50

j3d wrote:

... maybe should start a 3D mailing-list instead of being constantly off-topic in this thread

+1 (or a dedicated thread on this 3D forum so we could touch more osm contributors)

j3d · 2013-10-31 18:39:52

things-change: the first part I mentioned as an example how one could extract a likely rectangle subdivision from the outline. the other part is to handle the constraint that the smaller roof must connect with its ridge to the edge of the main roof - with this constraint height of both roofs(parts) could be inferred automatically as there seems to be only one solution

Last edited by j3d (2013-10-31 18:40:40)

j3d · 2013-10-31 18:54:43

cmif4: for my interpretation of an open dome no skeleton is needed. when the center is inside the inner ring:
1 connect lines from center to all points in outer ring
2 clip lines to building footprint
3 interpolate height along lines - this way the height is constant at the inner ring

I'll try that soon

Though when center is not inside the inner ring (or there are more inner rings) things get more complicated. Then one could build the triangle mesh for the rounded dome shape (as usual) and then insert the lines from the inner rings, put all that in a constrained delaunay triangulation and filter out the triangles inside the hole afterwards.

marek kleciak · 2013-11-06 12:19:53

cmif4,
look here:
http://www.openstreetmap.org/browse/way/79799800
There is no right interpretation in 3D.
A bug?

Regards,
Marek

cmif4 · 2013-11-06 12:28:19

As already said few post before, for now we don't handle roof:shape when polygons have holes.
The building part http://www.openstreetmap.org/browse/way/244031566 makes a hole on the outline when we merge building relation so outline roof shape is not handled.

I am currently working on buildings roof it should work in a few days, i already got dome, skillion and pyramidal working with holes (like the Pantheon in Roma).

!i! · 2013-11-06 13:00:55

f4 maps processing is outdated >20hrs, can you please fix it?

cmif4 · 2013-11-06 13:19:14

!i! wrote:

f4 maps processing is outdated >20hrs, can you please fix it?

We're already working on it, some osm data import broke our database.

Edit: check out this relation: http://www.openstreetmap.org/browse/relation/1980619 (prepare yourself for Internal Server Error )

Last edited by cmif4 (2013-11-06 13:36:28)

marek kleciak · 2013-11-06 14:11:40

i already got dome, skillion and pyramidal working with holes (like the Pantheon in Roma).

Congratulations!!

cmif4 · 2013-11-07 12:38:00

marek kleciak wrote:

Hello cmif4,
do You already have implemented roof type "gabled" for donut topology?

Got it working locally, check out the screen below.

j3d wrote:

I currently try to figure out to create proper 'gabled' roof types from the outline skeleton. It seems to be reasonable to extend only those ridges nodes to the outline which have two incoming edges that directly connect to the outline and where the edge pair must roughly have an angle of 90 degrees and point away from the ridge.

I'm working on straight skeleton aswell and i'm having some spectacular result on very complex shapes, but unfortunately it is very instable and it screws up very simple buildings that worked fine with the old method .

I'm also dealing with artifacts on buildings outlines incompletely used for rendering due to our "remove building:part from outline", skeleton goes outside and does not really match the 3d shape...

Anyway there is a sample of how it looks on some spots:

marek kleciak · 2013-11-07 14:45:05

cmif4,
please contact me.
I see, You´re very talented programmers. I teached architects geometry over 15 years.
There is an geometric approach for fixnix this problem, which is used by architect for construction of such building shapes.

cmif4 · 2013-11-07 15:18:37

You can contact me using the "Email" button under my avatar (it is not a trap )

j3d · 2013-11-07 15:39:46

cmif4: building heights of 'hotel des invalides' look wrong. did you mess with the map data?

BTW do you calculate the minimal distance through the skeleton to the outline to retrieve the relative heights? There should be a patch in sfcgal soon so that the skeleton contain the correct distances as z (or m) value :)

cmif4 · 2013-11-07 15:48:29

j3d wrote:

cmif4: building heights of 'hotel des invalides' look wrong. did you mess with the map data?

I split the building into 3 parts this morning as they do not have the same height http://binged.it/1cX4pJl

j3d wrote:

BTW do you calculate the minimal distance through the skeleton to the outline to retrieve the relative heights? There should be a patch in sfcgal soon so that the skeleton contain the correct distances as z (or m) value

Height is computed on client side (we do not add z in the database) so it can't be calculated during straight_skeleton process.

j3d · 2013-11-07 15:51:42

cmif4 wrote:

I split the building into 3 parts this morning as they do not have the same height http://binged.it/1cX4pJl

Oh indeed, you are right. I looked at some images of that building recently and was sure they were the same height..

kendzi · 2013-11-07 19:45:02

You plan to use skeleton algorithm only for roof tagged with roof:shape=hipped?

cmif4 · 2013-11-08 10:14:54

We plan to use it on

gabled, half-hipped, hipped, gambrel, mansard and round

maybe with some condition on the polygon (for example if shape is convex there is no point to build the skeleton, the old obox approach should work).

I started with hipped as skeleton does not need to be processed and can be used "as it" for triangulation.

I'm currently wondering how to handle the "roof:orientation=across" tag with skeleton on complicated shapes

marek kleciak · 2013-11-15 08:28:18

What is this?
http://map.f4-group.com/#lat=52.5199191 … hi=-13.178

We really need sphere as a point with height and radius for such purposes..

cmif4 · 2013-11-15 09:57:40

marek kleciak wrote:

What is this?
http://map.f4-group.com/#lat=52.5199191 … hi=-13.178

We really need sphere as a point with height and radius for such purposes..

We really need contributor to stop doing buildings inside buildings overlapping with buildings and all together inside a building relation (how about using part and shapes defines in S3DB )

marek kleciak · 2013-11-29 11:36:27

I see,
you´ve problems with rendering of roof:shape pyramidal with too many points.
See Heinrich Herz Tower in Hamburg / Germany
http://map.f4-group.com/#lat=53.5628755 … id=Hamburg

Btw.:

We really need contributor to stop doing buildings inside buildings overlapping with buildings and all together inside a building relation

Absolutely yes.

chris66 · 2013-11-29 12:01:53

Hi,
let me know how to tag boats:

and how to avoid that they are driving above the bridges.

cmif4 · 2013-11-29 12:02:28

There is no issue, i got the same rendering as Kendzi3D, the only difference is the outline that got a hole and is higher in F4Map due to the man_made=tower tag (our random height is based on building or man_made values).

Add "building:part=no" to the outline way and it will look what you expect.

Announcement

#176 2013-10-30 16:36:47

Re: Real time 3D map using WebGL

#177 2013-10-31 08:20:14

Re: Real time 3D map using WebGL

#178 2013-10-31 18:31:45

Re: Real time 3D map using WebGL

#179 2013-10-31 18:32:57

Re: Real time 3D map using WebGL

#180 2013-10-31 18:39:50

Re: Real time 3D map using WebGL

#181 2013-10-31 18:39:52

Re: Real time 3D map using WebGL

#182 2013-10-31 18:54:43

Re: Real time 3D map using WebGL

#183 2013-11-06 12:19:53

Re: Real time 3D map using WebGL

#184 2013-11-06 12:28:19

Re: Real time 3D map using WebGL

#185 2013-11-06 13:00:55

Re: Real time 3D map using WebGL

#186 2013-11-06 13:19:14

Re: Real time 3D map using WebGL

#187 2013-11-06 14:11:40

Re: Real time 3D map using WebGL

#188 2013-11-07 12:38:00

Re: Real time 3D map using WebGL

#189 2013-11-07 14:45:05

Re: Real time 3D map using WebGL

#190 2013-11-07 15:18:37

Re: Real time 3D map using WebGL

#191 2013-11-07 15:39:46

Re: Real time 3D map using WebGL

#192 2013-11-07 15:48:29

Re: Real time 3D map using WebGL

#193 2013-11-07 15:51:42

Re: Real time 3D map using WebGL

#194 2013-11-07 19:45:02

Re: Real time 3D map using WebGL

#195 2013-11-08 10:14:54

Re: Real time 3D map using WebGL

#196 2013-11-15 08:28:18

Re: Real time 3D map using WebGL

#197 2013-11-15 09:57:40

Re: Real time 3D map using WebGL

#198 2013-11-29 11:36:27

Re: Real time 3D map using WebGL

#199 2013-11-29 12:01:53

Re: Real time 3D map using WebGL

#200 2013-11-29 12:02:28

Re: Real time 3D map using WebGL

Board footer