How do I get the intersection region after boolean Union

[emil-peters]

Hi,

I was wondering how I can use the BooleanResultMapper to get the "interection" region of the 2 meshes I unioned

I have this but I can't find the correct settings. It either always returns the full self or other mesh,

    mr_self = mesh.toMrMesh()
    mr_other = other.toMrMesh()

    res_mapper = mr.BooleanResultMapper()
    result = mr.boolean(mr_self, mr_other, mr.BooleanOperation.Union, None, res_mapper)

    if result.errorString != "":
       exit()

    region_fbs = mr.FaceBitSet()
    region_fbs.resize(mr_other.topology.numValidFaces(), True)
    region_fbs = res_mapper.map(region_fbs, mr.BooleanResMapObj.B)

I know I can use this in DifferenceAB to find the fbs where the boolean cut was made.

I'd like to know this region to apply a remesh on the region, since booleaning flat planes often results in weird geometry.

[Grantim]

Hello!

There is newFaces function in mapper:

    mr_self = mesh.toMrMesh()
    mr_other = other.toMrMesh()

    res_mapper = mr.BooleanResultMapper()
    result = mr.boolean(mr_self, mr_other, mr.BooleanOperation.Union, None, res_mapper)

    if result.errorString != "":
       exit()

    region_fbs = res_mapper.newFaces()

[emil-peters]

Wow thanks, completely looked over that func

[emil-peters]

Hi @Grantim ,

Coming back to this issues. I get a lot of double free or corruption (!prev) with this approach.
Sometimes even during the booleaning
As often the new_faces_fbs is of a completely different size then the number of faces in the mesh.

While I don not have this issue with the approach below

    region_fbs = mr.FaceBitSet()
    region_fbs.resize(mr_other.topology.numValidFaces(), True)
    region_fbs = res_mapper.map(region_fbs, mr.BooleanResMapObj.B)

Any ideas why this might occur?

[Grantim]

I see, it looks like we fixed this in newer releases:
v3.0.2.183...v3.0.4.121#diff-08d395c625b872e10f09fd25a6b5739db053622ed8b03431af68d537e45606d1

Could you please try latest meshlib version?

[emil-peters]

Upgrading seems to fix the issue thanks!
Were any breaking changes introduced in between these versions?

[Grantim]

Great!

Were any breaking changes introduced in between these versions?

Not that I can remember, but you can check release notes:
https://github.com/MeshInspector/MeshLib/releases/tag/v3.0.3.233
https://github.com/MeshInspector/MeshLib/releases/tag/v3.0.4.121

[emil-peters]

Hi @Grantim

@hen upgrading from 3.0.2.183 to 3.0.3.233 an issue with voxelization seems to arise.

In 3.0.2.183 the voxelization seems to fill the entire mesh shell mesh with our configuration

While in 3.0.3.233 it is no longer the case. This makes it unusable for us

    def get_params_by_patch_size(
        patch_size, b_min, b_max, maxDistSq=None, signed=True
    ) -> mr.MeshToDistanceVolumeParams:
        "get parameters for the voxelization based on the patch size"
        patch_size = np.array(patch_size)[::-1]
        params = mr.MeshToDistanceVolumeParams()

        if signed:
            params.dist.signMode = mr.SignDetectionMode.HoleWindingRule
        else:
            params.dist.signMode = mr.SignDetectionMode.Unsigned

        mesh_dimentions = b_max - b_min
        dimention = np.array(patch_size, dtype=np.int_)
        params.vol.dimensions = mr.Vector3i(dimention[0], dimention[1], dimention[2])
        voxelSizes_numpy = mesh_dimentions / dimention
        voxelSizes = mr.Vector3f(
            voxelSizes_numpy[0], voxelSizes_numpy[1], voxelSizes_numpy[2]
        )
        params.vol.voxelSize = voxelSizes
        if maxDistSq is not None:
            params.dist.maxDistSq = maxDistSq
        else:
            params.dist.maxDistSq = 0.5
        return params


    params = get_params_by_patch_size(patch_size, b_min, b_max, maxDistSq=maxDistSq, signed=signed)

    # actual voxelization with meshlib
    params = get_params_by_patch_size(patch_size, b_min, b_max, maxDistSq=maxDistSq, signed=signed)
    volume = mr.meshToDistanceVolume(mr_mesh, params)
    npArray = mrn.getNumpy3Darray(volume)

    total_data = -npArray

Any ideas what might cause this?
As I'd like to upgrade the version for the fix of the booleans, but I can't as long as this is an issue 😢

[Grantim]

It looks like you need to set

params.dist.nullOutsideMinMax = False

[emil-peters]

Hi @Grantim

Sorry to open this again, but I am wondering if there is another bug introduced with the newest version? Currently on (3.0.4.121)

Using the 2 provided meshes I have the following issue with newFaces() as it returns a FaceBitSet larger than the available nb of faces. Causing all weird issues with dilations & remeshing

mr_self.topology.getValidFaces().size()
364
mr_other.topology.getValidFaces().size()
16
result.mesh.topology.getValidFaces().size()
622
res_mapper.newFaces().size()
673

    res_mapper = mr.BooleanResultMapper()
    result = mr.boolean(mr_self, mr_other, mr.BooleanOperation.Union, None, res_mapper)

    if result.errorString != "":
         exit()

   region_fbs = res_mapper.newFaces()

meshes.zip

[Grantim]

Hello @emil-peters

It is actually not a bug, while I agree that it might be a bit confusing:

1. One should not really consider size of BitSet as "available number of faces", only set bits matters.
2. In this function we resize bitset roughly for optimization (to avoid allocating more and more bits for it)
   

   MeshLib/source/MRMesh/MRBooleanOperation.cpp

   Lines 364 to 378 in f91a2c2

   	FaceBitSet BooleanResultMapper::newFaces() const
   	{
   	FaceBitSet res( std::max( maps[0].cut2newFaces.size(), maps[1].cut2newFaces.size() ) );
   	for ( const auto& map : maps )
   	{
   	for ( FaceId newF = 0_f; newF < map.cut2origin.size(); ++newF )
   	{
   	if ( newF == map.cut2origin[newF] || !map.cut2origin[newF].valid() )
   	continue;
   	if ( auto resF = map.cut2newFaces[newF] )
   	res.autoResizeSet( resF );
   	}
   	}
   	return res;
   	}

from meshlib import mrmeshpy as mm

mesh = mm.loadMesh(__file__+"/../mesh.stl")
other = mm.loadMesh(__file__+"/../other.stl")

mapper = mm.BooleanResultMapper()
res = mm.boolean(mesh,other,mm.BooleanOperation.Union,None,mapper).mesh
newFaces = mapper.newFaces()

print(mesh.topology.lastValidFace().get(),"/", mesh.topology.getValidFaces().size())
print(other.topology.lastValidFace().get(),"/", other.topology.getValidFaces().size())
print(res.topology.lastValidFace().get(),"/", res.topology.getValidFaces().size())
print(newFaces.find_last().get(),"/",newFaces.size())

363 / 364
15 / 16
621 / 622
621 / 673

You can see here that all "excessive" bits are not set.

---

Anyway you are right, we should change resize to reserve here, so size() will not be larger then actual mesh valid faces bitset size (while it could be smaller)

[Grantim]

#4480

[emil-peters]

Hmm I see, thanks for the clarification.