Marching cubes algorithm on own x,y,z data

[rajeshm71]

Hello all,
I am using marching cubes lewinar algorithm for surfaces reconstruction in 3d . I have data in the format of x,y,z coordinates. How to convert it into 3d volumetric data.

suppose i have
x=np.array([1,2,3])
y=np.array([4,5,6])
z=np.array([7,8,9])

Note: I have taken 3 points just for exapmle purpose

marchine cubes algorithm requires data in 3d numpy array. plz help

[jni]

@rajeshm71 can you tell us more specifically what you are aiming to do? The point of marching_cubes_lewiner is to go from volumetric data to triangles. If you already have triangles, there is no point doing the round trip via volumetric data. If you only have points and are looking to get a surface via triangulation, then you probably want to use scipy.spatial.Delaunay rather than marching cubes.

I'm going to close this issue because it is not a bug or feature request, but please feel free to continue the discussion below!

[rajeshm71]

Hi @jni Thanks for your reply. I have point cloud data as x,y,z coordinates in .ply format file. I am looking to get a surface by triangulation .I have tried with scipy.spatial.Delaunay but not getting proper results. so i was looking to reconstruct surface using marching cube algorithm. Is there any way by which i can convert pointcloud data to volumetric data and use marching cube algorithm for surface reconstruction in it.

[jni]

Yeah, Delaunay is actually not exactly what you need here, but I thought it might be good enough.

I think that raster + marching cubes will probably also not give you good results. I found a guide that seems to do what you want, without scikit-image (which is not designed for point clouds). Here's an attempt for the point-cloud -> volume -> mesh pipeline:

# input: z_coords, y_coords, x_coords
zint, yint, xint = [
    np.floor(coords).astype(int)
    for coords in [z_coords, y_coords, x_coords]
]
shape = tuple([np.max(intcoords) + 1
               for intcoords in [zint, yint, xint]])
image = np.zeros(shape)
image[zint, yint, xint] += 1

You can then apply marching cubes to that image. But, to reiterate, I suggest you instead follow the guide in this blog post:

https://towardsdatascience.com/5-step-guide-to-generate-3d-meshes-from-point-clouds-with-python-36bad397d8ba