Non BLocking Visualization in Open3d

Question

I have 2 different pointclouds stored in 2 different numpy arrays and I can visualize them combined in open3d. But what I want to do is that one pointcloud stays constant and the other moves in the z direction (that is updating the coordinates of numpy array) without closing the window.

Below is the code in which I pass these two numpy arrays (npa and npa_test) to visualize in open3d. Please note that in the initial lines I store numpy arrays to a .pcd files (since they were extracted from different types of files .asc which open3d doesnot recognize) and then read them using open3d read function. I want to update the one pointcloud geomtry by moving it in z direction and then visualizing it on the same window without closing it. This can be done by updating the z cordinates by 1 in for loop as mentioned below. Please let me know if you have a solution regarding this. Thank you.

def draw(npa,npa_test):

pcd = o3d.geometry.PointCloud()
pcd.points = o3d.utility.Vector3dVector(npa)
o3d.io.write_point_cloud("C:/Users/Junaid/Desktop/test.ply", pcd)
pcd_load = o3d.io.read_point_cloud("C:/Users/Junaid/Desktop/test.ply")

pcd1 = o3d.geometry.PointCloud()
pcd1.points = o3d.utility.Vector3dVector(npa_test)
o3d.io.write_point_cloud("C:/Users/Junaid/Desktop/test1.ply", pcd1)
pcd_load1 = o3d.io.read_point_cloud("C:/Users/Junaid/Desktop/test1.ply")

o3d.utility.set_verbosity_level(o3d.utility.VerbosityLevel.Debug)

vis = o3d.visualization.Visualizer()
vis.create_window()
vis.add_geometry(pcd_load)
vis.add_geometry(pcd_load1)

npa.tolist()

for i in range(len(npa)):

    npa[i][2] = npa[i][2] + 1

    npa=np.asarray(npa, dtype=np.float32)
    pcd = o3d.geometry.PointCloud()
    pcd.points = o3d.utility.Vector3dVector(npa)
    o3d.io.write_point_cloud("C:/Users/Junaid/Desktop/test.ply", pcd)
    pcd_load = o3d.io.read_point_cloud("C:/Users/Junaid/Desktop/test.ply")

    vis.update_geometry(pcd_load)

    vis.poll_events()
    vis.update_renderer()
vis.destroy_window()

Answer 1

I see that you're using npa[i][2] = npa[i][2] + 1 to modify the value which is overkill when you can just use Transformation Matrix to transform the geometry like pcd.transform(np.array([[1,0,0,0],[0,1,0,0],[0,0,1,z_val],[0,0,0,1]])).

Here's easy solution of moving sphere from z=0.5 until 15.0 with increment of 0.005 using open3d.

import open3d as o3d
import numpy as np

def create_transform_matrix_from_z(z):
    """ Return transform 4x4 transformation matrix given a Z value """
    result = np.identity(4)
    result[2,3] = z # Change the z
    
    return result

# Create Open3d visualization window
vis = o3d.visualization.Visualizer()
vis.create_window()

# create sphere geometry
sphere1 = o3d.geometry.TriangleMesh.create_sphere()
vis.add_geometry(sphere1)

# create coordinate frame
coordinate_frame = o3d.geometry.TriangleMesh.create_coordinate_frame()
vis.add_geometry(coordinate_frame)

prev_tf = None
for curr_z in np.arange(0.5, 15.0, 0.005):
    # return sphere1 to original position (0,0,0)
    if prev_tf is not None:
        sphere1.transform(np.linalg.inv(prev_tf))

    # transform bazed on curr_z tf
    curr_tf = create_transform_matrix_from_z(curr_z)
    sphere1.transform(curr_tf)

    prev_tf = curr_tf

    vis.update_geometry(sphere1)
    vis.poll_events()
    vis.update_renderer()