[vtkusers] Car Model

Mon Feb 9 05:22:21 EST 2009

Hi David,

I used Paraview to produce this first image but it was a trick: I applied
something I use as a basis for CAD/CAE conversion and automatic meshing.

Here is the procedure you should follow:

- open your VTU

- extract the surface

- ‘clean’ the surface (I used a 0.005 tolerance)

- extract the feature edges (only boundary edges)

You should get something like this:

Now, with Paraview, you can select some contours and use vtkDelaunay to fill
them 
 It will gave you a first result.

What I was suggesting instead is the use vtkStripper, to rebuild clean
contours from this, and then to use vtkTriangleFilter

vtkStripper holeStrips = new vtkStripper();

holeStrips.SetInput(featuresEdges.GetOutput());

holeStrips.Update(); 

vtkPolyData holePoly = new vtkPolyData();

holePoly.SetPoints(holeStrips.GetOutput().GetPoints());

holePoly.SetPolys(holeStrips.GetOutput().GetLines()); 

vtkTriangleFilter missingTriangles = new vtkTriangleFilter();

missingTriangles.SetInput(holePoly);

My guess is that you’ll have to do that more than one to fill every hole.

If it fills too much holes, maybe you should subset featuresEdges by region
with a connectivity filter before ‘stripping’ the whole set.

BTW, you could use that to remove the inner point instead of a convex hull /
threshold 

Try it with the ‘connectivity’ tool in paraview 

FYI : We used to drop a bag of flour or talcum powder on the object before
laser scanning. It helps the laser to detect the actual surface and to not
mess anymore with shiny or translucent surface ... 

Cordialement,

F.Danesi

-- 

Responsable Développement Informatique et Travail Collaboratif

Département DINCCS (Département Ingénierie Numérique, Conception
Collaborative et Simulation)

MICADO / DINCCS

Pôle de Haute Technologie, BP 90005

08005 Charleville-Mézières Cedex

Tel. : +33 (0) 3.24.41.69.55 / +33 (0) 6.62.76.13.32

Email : frederic.danesi at dinccs.com

Web : www.afmicado.com / www.dinccs.com

-----Message d'origine-----
De : David Doria [mailto:daviddoria at gmail.com] 
Envoyé : samedi 7 février 2009 15:21
À : Frederic Danesi
Objet : Re: Car Model

Fred -

I have several LiDAR scans from different angles of the same car that

have been registered to give the full surface rather than just 2.5D (a

scan from one position). The problem is that the laser goes through

the windows, so it's been tough to make a water tight model from the

resulting point cloud. I would like to know the idea/procedure so I

can create this type of mesh from other scans in the future as well. I

can use any combination of c++, paraview, and VTK. It may be easiest

to learn the steps if they can all be done in Paraview, because as far

as I know paraview completely uses VTK, so I can always go back and

figure out how to do whatever we do in Paraview in VTK.

Here are the raw points:

http://www.rpi.edu/~doriad/CarPoints.vtp

Another problem is that there are points inside the volume (the

headrests etc). I used a convex hull to threshold points that fell

only within some distance of the hull before I started trying to

create the mesh. Sound reasonable?

What was that "automatic filling procedure" that you used to produce

the image you sent me? It is very close to what I'm looking for -

there seem to still be some holes between the tires and the body of

the car - maybe they can be filled in the same way you filled in the

windows? How did you do that?

Thanks for all the help Fred!

David

On Fri, Feb 6, 2009 at 10:05 AM, Frederic Danesi

<frederic.danesi at dinccs.com> wrote:

> Hi Dave,

> 

> Using automatic filling method produce the following result. It does not

> sound really good to me 
 but it is a first step.

> 

> What exactly are you looking for ?

> 

> Is this a model you got by scanning a vehicle ? if so, maybe I can do
better

> with the points cloud or the STL directly.

> 

> 

> 

> You are working with paraview or VTK ? Do you just want to correct this

> model, or you're tried to write an application to do so ?

> 

> Fred.

> 

> 

> 

> 

> 

> 

> 

> 

> 

> Cordialement,

> 

> F.Danesi

> 

> --

> 

> Responsable Développement Informatique et Travail Collaboratif

> 

> Département DINCCS (Département Ingénierie Numérique, Conception

> Collaborative et Simulation)

> 

> MICADO / DINCCS

> 

> Pôle de Haute Technologie, BP 90005

> 

> 08005 Charleville-Mézières Cedex

> 

> Tel. : +33 (0) 3.24.41.69.55 / +33 (0) 6.62.76.13.32

> 

> Email : frederic.danesi at dinccs.com

> 

> Web : www.afmicado.com / www.dinccs.com

> 

> 

> 

> 

> 

> -----Message d'origine-----

> De : daviddoria at gmail.com [mailto:daviddoria at gmail.com]

> Envoyé : vendredi 6 février 2009 15:20

> À : frederic.danesi at dinccs.com

> Objet : Car Model

> 

> 

> 

> Hi Fred- thanks for the offer! I'll take you up on it only because I've
been

> trying to do this off and on for a couple of months now haha.

> 

> Here is a link to the model

> 

> http://rpi.edu/~doriad/Car.vtp

> 

> 

> 

> I would like to get a watertight, non-convex (ie still looks very much
like

> the car) mesh from this. I tried to follow your instructions: I used the

> "extract edges filter" and it seemed to work. Then I tried to do "Extract

> cells by region" but I didn't know what I was doing so I just left it

> default.  Then I did "Extract surface" but nothing happened.

> 

> 

> 

> Any tips would be great!

> 

> 

> 

> Thanks,

> 

> 

> 

> Dave

> 

> Ce message entrant est certifié sans virus connu.

> 

> Analyse effectuée par AVG - www.avg.fr

> 

> Version: 8.0.233 / Base de données virale: 270.10.18/1936 - Date: 02/05/09

> 11:34:00

-- 

Thanks,

David

Ce message entrant est certifié sans virus connu.

Analyse effectuée par AVG - www.avg.fr 

Version: 8.0.233 / Base de données virale: 270.10.19/1939 - Date: 02/06/09
11:31:00

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20090209/2b1c4fdb/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 9948 bytes
Desc: not available
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20090209/2b1c4fdb/attachment.png>