I would like to share how to used dynamo to registration cloudpoint (if you search for patern of point cloud into a big point cloud) this post will help you to detect point cloud patern into big pointcloud
i use unsupervised machen learning Open3d packedge.
http://www.open3d.org/docs/release/tutorial/pipelines/icp_registration.html#Point-to-point-ICP
1- select a patern point cloud
(rail track patern)
2-select target point cloud
3- run Dynamo to detect the this patern into the target point cloud
in this example i used it to detect the rail track and convert it to dynamo node
"""
this code write by Ramiz Mohareb
enramiz@yahoo.com
07-02-2022
Important! This code will only work with Dynamo versions
that are able to run cPython3 and have the following additional
python packages installed.
Packages required:
numpy
"""
import random
import numpy as np
import clr
import open3d
#libraries used
import numpy as np
import open3d as o3d
clr.AddReference('ProtoGeometry')
from Autodesk.DesignScript.Geometry import *
import matplotlib.pyplot as plt
filepath=IN[0]
import numpy as np
import open3d as o3d
filepath=IN[0]
segment_models={}
segments={}
inliers={}
inlier_cloud={}
Outt=[]
Outt2=[]
Outt3=[]
dec_meshL=[]
max_plane_idx=1
point_cloud= np.loadtxt(filepath,skiprows=1)
#Format to open3d usable objects
pcd = o3d.geometry.PointCloud()
pcd.points = o3d.utility.Vector3dVector(point_cloud[:,:3])
#pcd.colors = o3d.utility.Vector3dVector(point_cloud[:,3:6]/255)
#pcd.normals = o3d.utility.Vector3dVector(point_cloud[:,6:9])
#source=pcd.voxel_down_sample(voxel_size=0.01)
OUT =pcd
"""
this code write by Ramiz Mohareb
enramiz@yahoo.com
07-02-2022
Important! This code will only work with Dynamo versions
that are able to run cPython3 and have the following additional
python packages installed.
Packages required:
numpy
"""
import random
import numpy as np
import clr
import open3d
import copy
#libraries used
import numpy as np
import open3d as o3d
import trimesh
clr.AddReference('ProtoGeometry')
from Autodesk.DesignScript.Geometry import *
import matplotlib.pyplot as plt
Outt=[]
# ICP algortihm requires a good initial allignment to converge efficiently.
trans_init = np.asarray([[1, 0, 0, 0],
[0, 1, 0, 0],
[0, 0, 1, 0], [0, 0, 0, 1]])
source1 = IN[0]
target1 = IN[1]
source2 = copy.deepcopy(source1).translate((-3572567.317, -5363546.02427989, -502.8697))
target = copy.deepcopy(target1).translate((-3572567.317, -5363546.02427989, -502.8697))
#source = copy.deepcopy(source2).translate((30,10 ,10))
source=source2
New_target=target
New_source=source
criteria = o3d.pipelines.registration.ICPConvergenceCriteria(max_iteration=10000)
# Down-sampling voxel-size. If voxel_size < 0, original scale is used.
voxel_size = -1
reg_p2p = o3d.pipelines.registration.registration_icp(source, New_target,1000, trans_init,o3d.pipelines.registration.TransformationEstimationPointToPoint(),criteria)
#Outt.append(np.asarray(reg_p2p.transformation).tolist())
#draw_registration_result(source, target, reg_p2p.transformation)
New_source = copy.deepcopy(source).transform(reg_p2p.transformation)
dists = target.compute_point_cloud_distance(New_source)
dists = np.asarray(dists)
ind = np.where(dists > 0.1)[0]
New_target = New_target.select_by_index(ind)
##source = copy.deepcopy(New_source).translate((30,30 ,30))
OUT =New_source,New_target#,np.asarray(New_target.points).tolist(),np.asarray(New_source.points).tolist()
"""
this code write by Ramiz Mohareb
enramiz@yahoo.com
07-02-2022
Important! This code will only work with Dynamo versions
that are able to run cPython3 and have the following additional
python packages installed.
Packages required:
numpy
"""
import random
import numpy as np
import clr
import open3d
import copy
#libraries used
import numpy as np
import open3d as o3d
import trimesh
clr.AddReference('ProtoGeometry')
from Autodesk.DesignScript.Geometry import *
import matplotlib.pyplot as plt
OUTT=[]
OUTT2=[]
Fitt=[]
# ICP algortihm requires a good initial allignment to converge efficiently.
trans_init = np.asarray([[1, 0, 0, 0],
[0, 1, 0, 0],
[0, 0, 1, 0], [0, 0, 0, 1]])
source = IN[0]
target = IN[1]
New_source={}
TO={}
New_target={}
New_target[0]=target
New_source[0]=source
criteria = o3d.pipelines.registration.ICPConvergenceCriteria(max_iteration=100000)
# Down-sampling voxel-size. If voxel_size < 0, original scale is used.
voxel_size = -1
for i in range(10):
reg_p2p = o3d.pipelines.registration.registration_icp(New_source[i], New_target[i],10, trans_init,o3d.pipelines.registration.TransformationEstimationPointToPoint(),criteria)
#Outt.append(np.asarray(reg_p2p.transformation).tolist())
#draw_registration_result(source, target, reg_p2p.transformation)
New_source[i+1] = copy.deepcopy(New_source[i]).transform(reg_p2p.transformation)
dists = New_target[i].compute_point_cloud_distance(New_source[i+1])
dists = np.asarray(dists)
ind = np.where(dists > 0.1)[0]
ind2 = np.where(dists < 0.1)[0]
New_target[i+1] = New_target[i].select_by_index(ind)
OUTT.append(New_source[i+1])
OUTT2.append(New_target[i].select_by_index(ind2))
evaluation = o3d.pipelines.registration.evaluate_registration(
New_source[i+1], New_target[i+1] , 0.2, trans_init)
Fitt.append(evaluation)
OUT =OUTT,Fitt,OUTT2#,#New_target,New_source #,np.asarray(New_source.points).tolist(),np.asarray(New_target.points).tolist()
UPDATE
also you can convert the point cloud to revit element mesh
"""
this code write by Ramiz Mohareb
enramiz@yahoo.com
07-02-2022
Important! This code will only work with Dynamo versions
that are able to run cPython3 and have the following additional
python packages installed.
Packages required:
numpy
"""
import random
import numpy as np
import clr
import open3d
#libraries used
import numpy as np
import open3d as o3d
clr.AddReference('ProtoGeometry')
from Autodesk.DesignScript.Geometry import *
import matplotlib.pyplot as plt
import numpy as np
import open3d as o3d
segment_models={}
segments={}
inliers={}
inlier_cloud={}
Outt=[]
Outt2=[]
Outt3=[]
dec_meshL=[]
max_plane_idx=1
downpcd=IN[0]
distances = downpcd.compute_nearest_neighbor_distance()
avg_dist = np.mean(distances)
radius = 1 * avg_dist
#computing the mehs
bpa_mesh = o3d.geometry.TriangleMesh.create_from_point_cloud_ball_pivoting(downpcd,o3d.utility.DoubleVector([1*radius, radius*2]))
#dec_mesh1 = bpa_mesh.filter_smooth_simple(number_of_iterations=1)
dec_mesh1 = bpa_mesh.filter_smooth_taubin(number_of_iterations=10)
dec_mesh3=dec_mesh1
dec_mesh3.remove_degenerate_triangles()
dec_mesh3.remove_duplicated_triangles()
dec_mesh3.remove_duplicated_vertices()
dec_mesh3.remove_non_manifold_edges()
# Outt.append(np.asarray(segments[i].Points).tolist())
Outt2.append(np.asarray(dec_mesh3.triangles).tolist())
Outt3.append(np.asarray(dec_mesh3.vertices).tolist())
OUT = Outt2,Outt3
Point_Cloud_ICP_Track4.dyn (37.8 KB)
