Visualise a 2D image (from array of x, y, z and data points)

[kinverarity1]

Description

I'm a bit stuck on how to view a 2D image in a 3D scene. The 2D image is a conductivity model inverted from airborne EM data, collected along a flight line.

I have an array of x, y, z and data points. The x coordinates are not constant, but broadly similar (the flight lines are north-south). The y coordinates are not at a constant spacing, but roughly so. The z coordinates have a similar spacing, but a different starting point, as the elevation changes from north to south. There are the same number of z coordinates for each set of x, y coordinates.

Example code:

import pyvista as pv
import numpy as np
import pandas as pd

df = pd.read_csv("example_flightline.csv")
array = df.sort_values(["elev", "northing", "easting"]).values
ny = len(df.northing.unique())
nz = len(df[(df.easting == soundings.easting[0]) & (df.northing == soundings.northing[0])])
mesh = pv.StructuredGrid()
mesh.points = array[:, :3]
mesh.dimensions = [232, 30, 1]

Full example notebook here: https://github.com/kinverarity1/paraview-2d-viz/blob/master/trying%20to%20create%20structuredgrid.ipynb

Example Data

example_flightline.zip

[kinverarity1]

@banesullivan has very kindly helped me out with this!

So it looks like your points aren’t ordered the way the structured grid needs them. I got it to work by changing where you sort the points from the dataframe:

array = df.sort_values(["northing", "easting", "elev"]).values

then change the dimensions to (1, 30, 232):

dims = (1, 30, 232)
mesh = pv.StructuredGrid()
mesh.points = array[:, :3]
assert np.product(dims) == len(mesh.points)
mesh.dimensions = dims

mesh['log10(cond)'] = np.log10(array[:, -1])

mesh.plot( show_edges=False, show_grid=False, cpos="yz", notebook=False)

Why have the dimensions as (1, 30, 232)? That’s x, z, y - which is how we want to structure the points for this particular data. Your x is constant so ignore that and have it first, then you iterate over the points along the z axis so you will come across 30 nodes before moving to the next sounding which you have 232 of… the dimensions are less about how many nodes are on each axes and more about what is the ordering/structure of the points array because all VTK is doing is creating an indexing array to make the quads between the nodes which doesn’t care about where the nodes are located, just their order with respect to each other.

[banesullivan]

Thanks for posting here @kinverarity1!

This is related to pyvista/pyvista#318

[banesullivan]

And for completeness, the resulting grid looks like:

...
p = pv.BackgroundPlotter()
p.add_mesh(mesh, show_edges=True, nan_opacity=0.)