VRML - data vziualization and rendering

Author

data vziualization and rendering

LightningBit.News

2003-12-17, 9:42 am

Ok,

this question may sound strange, or out of scope for these newsgroups, but I
would not know where to start otherwise.

I've a collection of data, which can represented in a record as follow :

Record :
field1 xxxx,xxxx,xxxx,yyyy,zzzz
field2 xxxx,xxxx,xxxx,yyyy,zzzz

where xxxx,xxxx,xxxx can be coordinates, and yyyy and zzzz can be properties
or so

now we would like to represent that data in a visual form, kind of a visual,
3D map

each record has sets of 2 coordinates (=the relation), and the basic idea is
that each set of coordinates becomes a "dot", and a record of the 2
indicates a link between the 2.

so e.g. 2 dots and a line between the 2

the data collection can grow very large hundreds of thousands of records
a specific relation can only exist once, but from 1 set of coordinates can
start tens or hundreds of relations

as in one dot, will have many connections to other dots....

optionally, we woult use the extra data (yyyy, zzzz) to indicate e.g. the
color of the dot and the type (sphere, cube, ...)

and so on... I think you get the idea (I hope you do..... not easy to
explain)

ok so ... the first part of the question : what would be a good tool to do
this?
it looks like a tool like PovRay is interesting, feeding it with coordinates
and such

the second part is... is there a way to render this "live" this means, each
time it finds more data, build on to the existing scene, adding "the dots
and the lines"

Ok, that's it......

any takers? or anyone who can give some good hints??

thanks

O.

Bill Angel

2003-12-17, 9:42 am

Hi:

Here's one idea...

You state that the total data set may contain hundreds of thousands of records.

One approach would be similar to what is done when utilizing VRML to display a terrain map.

When the viewer is looking at the total dataset in its entirety, it would be like looking a terrain map
from a high altitude. The scene would not display all the details of the terrain, but just a low resolution
depiction,. As the user "zoomed" in to one section of the terrain, the low resolution depiction would be
replaced with one of higher resolution, but only for that section of the terrain map that currently would be visible.

As the user "zoomed in" further, the resolution would increase again to the "high magnification" setting, which
would correspond to the depiction of the individual nodes and connections in your dataset. These depictions
at different magnifications could be computed in real time, i.e as they are needed This would make the depiction
of the dataset at each level of detail (as the viewer "zooms in") to be as current as the most recent data in your database.

--- Bill Angel

"LightningBit.News" <news@$NOSPAM*Lightningibit.com> wrote in message news:bUrDb.80043$mG7.3353745@phobos.telenet-ops.be...

quote:

> Ok,
>
> this question may sound strange, or out of scope for these newsgroups, but I
> would not know where to start otherwise.
>
> I've a collection of data, which can represented in a record as follow :
>
> Record :
> field1 xxxx,xxxx,xxxx,yyyy,zzzz
> field2 xxxx,xxxx,xxxx,yyyy,zzzz
>
> where xxxx,xxxx,xxxx can be coordinates, and yyyy and zzzz can be properties
> or so
>
> now we would like to represent that data in a visual form, kind of a visual,
> 3D map
>
> each record has sets of 2 coordinates (=the relation), and the basic idea is
> that each set of coordinates becomes a "dot", and a record of the 2
> indicates a link between the 2.
>
> so e.g. 2 dots and a line between the 2
>
> the data collection can grow very large hundreds of thousands of records
> a specific relation can only exist once, but from 1 set of coordinates can
> start tens or hundreds of relations
>
> as in one dot, will have many connections to other dots....
>
> optionally, we woult use the extra data (yyyy, zzzz) to indicate e.g. the
> color of the dot and the type (sphere, cube, ...)
>
> and so on... I think you get the idea (I hope you do..... not easy to
> explain)
>
>
> ok so ... the first part of the question : what would be a good tool to do
> this?
> it looks like a tool like PovRay is interesting, feeding it with coordinates
> and such
>
> the second part is... is there a way to render this "live" this means, each
> time it finds more data, build on to the existing scene, adding "the dots
> and the lines"
>
>
> Ok, that's it......
>
> any takers? or anyone who can give some good hints??
>
> thanks
>
> O.
>
>
>

Joerg Scheurich aka MUFTI

2003-12-17, 9:42 am

> I've a collection of data, which can represented in a record as follow :
....

quote:

> the data collection can grow very large hundreds of thousands of records

....

quote:

> it looks like a tool like PovRay is interesting, feeding it with coordinates
> and such

....

quote:

> the second part is... is there a way to render this "live" this means

With that amount of data, you should not search a tool.

There is simply no such tool to view this data on a 1000x1000 computer
screen. If you have information with 10 pixels in your record, one
hundreds of thousands of records would fill the screen completely.

3D makes the screen not bigger, only deeper 8-(

BTW: plotterpaper has a bigger resolution than a computer screen 8-)

So you should search a algorithm, to hide the currently uninteresting parts
of the dataset.
For example, VRML has a Level Of Detail (LOD) node, that allows to exchange
a object far away with a simplified version of this object.

But with that amount of data, you would get a horrible big VRML file.
If you want to use VRML, it would be better, to do the filtering in a
external program and send only the currently interesting data to the
visualisation.
In VRML this can be done with the java EAI (external authoring interface).

so long
MUFTI
--
6 von 5 Datei(en) erfolgreich empfangen.
Statusmeldung von Leonardo

Herbert Stocker

2003-12-17, 9:42 am

LightningBit.News wrote:

quote:

> Ok,
>
> this question may sound strange, or out of scope for these newsgroups, but I
> would not know where to start otherwise.

No, it's not.

VRML _is_ a tool for visualizing scintific things.

Since you ask for PovRay, you need to understand that
if you use such a renderer you get still images or
a movie.

With VRML you get some live 3D. The scene is rendered in
realtime and you fly throug it or rotate an object.
It is also possible to have interaction with the scene, e.g.
by clicking on an object or draging it.
Think of it like of a flash film, but in 3D.

VRML is actually a language for 3D content, like HTML is for
text. There are various browsers, and if you use windows you
should check them out as they vary in rendering speed.

VRML is a descriptive language for 3D content. And it allows
to integrate JavaScript for programmable code.
One of the functions available in JavaScript is createVrmlFromString(.)
It allows you to create 3D objects on the fly.
Just say:

MainGroup.addChildren= Browser.createVrmlFromString(
'Shape { appearance Apperance { } geometry Box { size 1 3 9 } }');

and you have added a box of siz 1 x 3 x 9 units and with the default
properties for the appearance to the scene.
(Provided MainGroup is declared as necessary.)

VRML is not limitted to JavaScript. You can also use java if you
want, or PERL on linux/unix or, i think Visual Basic and JScript
on Windows.

quote:

>
> I've a collection of data, which can represented in a record as follow :
>
> Record :
> field1 xxxx,xxxx,xxxx,yyyy,zzzz
> field2 xxxx,xxxx,xxxx,yyyy,zzzz
>
> where xxxx,xxxx,xxxx can be coordinates, and yyyy and zzzz can be properties
> or so
>
> now we would like to represent that data in a visual form, kind of a visual,
> 3D map

Recently i have made a VRML application with Contact, which
has some data stored in an X3D file (an XMLized version of VRML (Umm))
and the VRML scene reloads this file from time to time and
updates the 3D scene according to the data read from the file.

Ah, apropos XML:
Using MS XML from Internet Explorer, and Contact you can convert an
XML file using XSLT and feed it directly into the VRML browser.
It's also possible to use DOM to modify the XML file before
converting it.

quote:

>
> each record has sets of 2 coordinates (=the relation), and the basic idea is
> that each set of coordinates becomes a "dot", and a record of the 2
> indicates a link between the 2.
>
> so e.g. 2 dots and a line between the 2
>
> the data collection can grow very large hundreds of thousands of records
> a specific relation can only exist once, but from 1 set of coordinates can
> start tens or hundreds of relations

You would have to try if this amount of data can be rendered live.

Probably you need to use some extensions of browsers for render
optimization, that go beyond what the VRML standard describes.
I think OctTree culling is usefull in this case.

quote:

>
> as in one dot, will have many connections to other dots....
>
> optionally, we woult use the extra data (yyyy, zzzz) to indicate e.g. the
> color of the dot and the type (sphere, cube, ...)

100 000 records with each connecting 2 spheres. Quite a lot.
But such an image would not help a human if all the spheres
were visible at the same time. (Unless the shape of the cloud
they form is interesting.)
Therefore you need to structure the scene anyway to show only
the interesting objects that are close to the viewing position.
Then the rendering time should not be a problem anymore.

quote:

>
> <...>
>

quote:

> Ok, that's it......
>
> any takers? or anyone who can give some good hints??
>

You can try BS Contact VRML from http://www.bitmanagement.de .
It supports a lot of extensions over VRML, you can view them
at http://www.bitmanagement.de/developer . See "Feature Labs"
and "Node Extension Demos".

For optimizing large scenes, Contac supports the following
techiques:
Cells & Portals
BSP / Quad / OctTree
Occlusion Culling

Herbert

--
Herbert Stocker (aka hersto)
http://www.hersto.de
www.bitmanagement.de