Show mesh registration features in slam

# Authors: Guillaume Auzias <guillaume.auzias@univ-amu.fr>

# License: BSD (3-clause)
# sphinx_gallery_thumbnail_number = 2

importation of slam modules

import slam.plot as splt
import slam.io as sio
import numpy as np
import trimesh

loading an two unregistered meshes

mesh_1 = sio.load_mesh("../examples/data/example_mesh.gii")
mesh_2 = sio.load_mesh("../examples/data/example_mesh_2.gii")

plot them to check the misalignment

joint_mesh = mesh_1 + mesh_2
joint_tex = np.ones((joint_mesh.vertices.shape[0],))
joint_tex[: mesh_1.vertices.shape[0]] = 10
visb_sc = splt.visbrain_plot(
    mesh=joint_mesh, tex=joint_tex, caption="before registration"
)
visb_sc.preview()

compute ICP registration this functionnality requires to install the optional package rtree

transf_mat, cost = trimesh.registration.mesh_other(
    mesh_1, mesh_2, samples=500, scale=False, icp_first=10, icp_final=100
)
print(transf_mat)
print(cost)

Out:

[[-0.59202764 -0.09421463  0.8003917  62.85878541]
 [ 0.09168714  0.97882135  0.18303614 53.43305704]
 [-0.80068517  0.18174808 -0.57085103 49.62517482]
 [ 0.          0.          0.          1.        ]]
0.006163062935171203

apply the estimated rigid transformation to the mesh

mesh_1.apply_transform(transf_mat)

Out:

<trimesh.Trimesh(vertices.shape=(2328, 3), faces.shape=(4652, 3))>

plot the two meshes to check they are now aligned

joint_mesh = mesh_1 + mesh_2
visb_sc = splt.visbrain_plot(
    mesh=joint_mesh, tex=joint_tex, caption="after registration"
)
visb_sc.preview()