Reversed triangles in vtkPolyData created from points and cells

Development
1

Hi,
I want to create a mesh vtkPolyData from points and triangles. Here’s my mesh topology.
image

Given the topology, I can create vtkPolyData from the points and triangles.
suedocode:

    auto points = vtkSmartPointer<vtkPoints>::New();
    for(const auto& point: inputPoints)
        points->InsertNextPoint(point.GetData());

    auto cells = vtkSmartPointer<vtkCellArray>::New();
    cells->InsertNextCell(GetTriangle(0, 1, 2));
    cells->InsertNextCell(GetTriangle(2, 1, 3));
    cells->InsertNextCell(GetTriangle(3, 1, 4));
    cells->InsertNextCell(GetTriangle(4, 2, 3));
    cells->InsertNextCell(GetTriangle(4, 0, 2));

    auto mesh = vtkSmartPointer<vtkPolyData>::New();
    mesh->SetPoints(points);
    mesh->SetPolys(cells);

where GetTriangle(int, int, int) returns a vtkSmartPointer<vtkTriangle>.
When I run the program and export the result vtkPolyData to a .stl file, it looks like this

image
image1025×717 9.62 KB

the 2 pink triangles are the ones with reversed normal.
Initially I thought this is because the vertices order of these triangles are wrong, but even if I changed the vertices order, the result is still the same.

Then I thought maybe my code is wrong. I checked the code by keeping the vertices the same, and get rid of triangle (4,2,3) and (4,0,2), as shown in picture

image
image969×749 17.1 KB

the result would be a valid mesh without any reversed triangles

image
image1022×745 7.98 KB

Why in the first case, there are reversed triangles?

Thanks

2

Hello,

If the indexes of your cartoon match that of the points in code, then all your faces are counter-clockwise when seen from the inside of the pyramid, which would result in consistent normals (either all pointing to inside or all to the outside). I’d do a double-check on the part of the code where you set the vertexes. They may be in an order different from the cartoon’s.

best,

PC

3

Hi Paulo,
Thanks for reply. As I said, initially I thought it’s the order problem. ie

 cells->InsertNextCell(GetTriangle(4, 2, 3));
 cells->InsertNextCell(GetTriangle(4, 0, 2));

these 2 lines produce reversed triangles. The vertices are in counter-clockwise order. So if I reverse the vertices of these 2 triangles

 cells->InsertNextCell(GetTriangle(3, 2, 4));
 cells->InsertNextCell(GetTriangle(2, 0, 4));

the vertices are in clockwise order. But this produces the exact same result.

4

Hello,

Try to recompute the normals with the vtkPolyDataNormals filter:

    // Compute normals
    vtkSmartPointer<vtkPolyDataNormals> normals = vtkSmartPointer<vtkPolyDataNormals>::New();
    normals->SetInputData(mesh);
    normals->ComputePointNormalsOn();
    normals->ComputeCellNormalsOn();
    normals->Update();

Use the object returned by normals->GetOutput() or normals->GetOutputPort() depending on whatever the remainder of your pipeline code is.

best,

PC

5

Hi Paulo,
Thanks for the help. Unfortunately, even with your code, the result mesh exported to .stl file still looks the same as before. However, I did find something interesting: If I change export file extension to PLY instead of STL, I get the correct result with my original code.

image
image2076×726 28.5 KB

I think there might be a bug with vtkSTLWriter?
Here’s my code for reproducing the result above.

#include <vtkVector.h>
#include <vtkTriangle.h>
#include <vtkPolyData.h>
#include <vtkPolyDataNormals.h>
#include <vtkSTLWriter.h>

vtkSmartPointer<vtkTriangle> GetTriangle(const int vid0, const int vid1, const int vid2)
{
    auto triangle = vtkSmartPointer<vtkTriangle>::New();

    triangle->GetPointIds()->SetId(0, vid0);
    triangle->GetPointIds()->SetId(1, vid1);
    triangle->GetPointIds()->SetId(2, vid2);

    return triangle;
}

void Test()
{
    std::vector<vtkVector3d> inputPoints(5);
    inputPoints[0] = vtkVector3d(11.97, 0, 4.75);
    inputPoints[1] = vtkVector3d(1.83, 0, 20.15);
    inputPoints[2] = vtkVector3d(9.38, 0, 11.95);
    inputPoints[3] = vtkVector3d(16, 0, 15);
    inputPoints[4] = vtkVector3d(22.73, 0, 17);

    auto points = vtkSmartPointer<vtkPoints>::New();
    for(const auto& point: inputPoints)
        points->InsertNextPoint(point.GetData());

    auto cells = vtkSmartPointer<vtkCellArray>::New();
    cells->InsertNextCell(GetTriangle(0, 1, 2));
    cells->InsertNextCell(GetTriangle(2, 1, 3));
    cells->InsertNextCell(GetTriangle(3, 1, 4));
    cells->InsertNextCell(GetTriangle(4, 2, 3));
    cells->InsertNextCell(GetTriangle(4, 0, 2));

    auto mesh = vtkSmartPointer<vtkPolyData>::New();
    mesh->SetPoints(points);
    mesh->SetPolys(cells);

    //const auto outDir = "test.stl";
    auto writer = vtkSmartPointer<vtkSTLWriter>::New();
    writer->SetFileName(outDir);
    writer->SetInputData(mesh);
    writer->Write();
}
6

Hello,

If you think it’s a bug, please, report it here: Issues · VTK / VTK · GitLab (kitware.com).

regards,

PC

7

Hi Paulo,
I tried to submit an issue, but GitLab keeps poping up

The form contains the following error:

* Your issue has been recognized as spam. Please, change the content to proceed.

What should I do to get rid of this error message?

8

Greetings,

Dunno. I’ve never had such message before. I don’k know what you’re trying to post, but do as told: review the message. Describe the problem with the like you did here.

best,

PC

9

What is your gitlab username ?

10

it’s veggiegJie

11

You should now be able to post on gitlab.

12

Thanks, I created the issue successfully.
May I know what’s the cause of the error message “Your issue has been recognized as spam. Please, change the content to proceed.”?

13

Our antispam system is a bit agressive with new users.

14

I see. Thanks.

15

https://gitlab.kitware.com/vtk/vtk/-/issues/19404