-
Notifications
You must be signed in to change notification settings - Fork 0
/
4_Generating_3D_Mesh.py
155 lines (129 loc) · 5.82 KB
/
4_Generating_3D_Mesh.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
# Import necessary libraries
import numpy as np
import matplotlib.pyplot as plt
import open3d as o3d
import laspy
import pandas as pd
import alphashape as ash
import geopandas as gpd
import shapely as sh
# 2. Data Profiling
# Load the neighborhood point cloud (ensure lazrs or laszip is installed for .laz files)
try:
las = laspy.read(r'data/neighborhood.laz')
except laspy.LaspyException as e:
print(f"Error reading the LAS file: {e}")
# Display unique classifications in the dataset
print("Unique classifications in the dataset:", np.unique(las.classification))
# List available dimensions in the LAS file
print("Available dimensions in the LAS file:", [dimension.name for dimension in las.point_format.dimensions])
# Check and display CRS information if available
if len(las.vlrs) > 2:
crs = las.vlrs[2].string
print("Coordinate Reference System (CRS):", crs)
else:
print("No CRS information available in VLRs.")
# 3. Data Pre-Processing
# 3.1. Building points initialization
# Create a mask to filter points classified as buildings (classification code 6)
pts_mask = las.classification == 6
xyz_t = np.vstack((las.x[pts_mask], las.y[pts_mask], las.z[pts_mask]))
pcd_o3d = o3d.geometry.PointCloud()
pcd_o3d.points = o3d.utility.Vector3dVector(xyz_t.transpose())
pcd_center = pcd_o3d.get_center()
pcd_o3d.translate(-pcd_center)
# o3d.visualization.draw_geometries([pcd_o3d])
# Ground Plane
pts_mask = las.classification == 2
xyz_t = np.vstack((las.x[pts_mask], las.y[pts_mask], las.z[pts_mask]))
ground_pts = o3d.geometry.PointCloud()
ground_pts.points = o3d.utility.Vector3dVector(xyz_t.transpose())
ground_pts.translate(-pcd_center)
# o3d.visualization.draw_geometries([ground_pts])
# Get Average Distance between building points
nn_distance = np.mean(pcd_o3d.compute_nearest_neighbor_distance())
print("Average distance between building points:", nn_distance)
# 4. Unsupervised Segmentation
epsilon = 2
min_cluster_size = 100
labels = np.array(pcd_o3d.cluster_dbscan(eps=epsilon, min_points=min_cluster_size, print_progress=True))
max_label = labels.max()
print(f"point cloud has {max_label + 1} clusters")
colors = plt.get_cmap("tab20")(labels / (max_label if max_label > 0 else 1))
colors[labels < 0] = 0
pcd_o3d.colors = o3d.utility.Vector3dVector(colors[:, :3])
# o3d.visualization.draw_geometries([pcd_o3d])
# 5. Selecting a segment
sel = 1
segment = pcd_o3d.select_by_index(np.where(labels == sel)[0])
# o3d.visualization.draw_geometries([segment])
# 6. Extracting the segment
points_2D = np.asarray(segment.points)[:, :2]
building_vector = ash.alphashape(points_2D, alpha=0.5)
# 7. Store in a geodataframe the 2D polygon
building_gdf = gpd.GeoDataFrame(geometry=[building_vector], crs='EPSG:26910')
print(building_gdf.head(1))
# 7.1 Computing semantics and attributes
# Height of the building
altitude = np.asarray(segment.points)[:, 2]+pcd_center[2]
height_test = np.max(altitude) - np.min(altitude)
print("Height of the building (Initial):", height_test)
# Define Ground Plane
query_point = segment.get_center()
query_point[2] = segment.get_min_bound()[2]
pcd_tree = o3d.geometry.KDTreeFlann(ground_pts)
[k, idx, _] = pcd_tree.search_knn_vector_3d(query_point, 200)
sample = ground_pts.select_by_index(idx, invert=False)
sample.paint_uniform_color([0.5, 0.5, 0.5])
# o3d.visualization.draw_geometries([sample, ground_pts])
ground_zero = sample.get_center()[2]
height = segment.get_max_bound()[2] - ground_zero
print("Height of the building (Final):", height)
print("Difference in height:", height - height_test)
# 7.2 Computing parameters
building_gdf[['id']] = sel
building_gdf[['height']] = height
building_gdf[['area']] = building_vector.area
building_gdf[['perimeter']] = building_vector.length
building_gdf[['local_cx', 'local_cy', 'local_cz']] = np.asarray([building_vector.centroid.x, building_vector.centroid.y, ground_zero])
building_gdf[['transl_x', 'transl_y', 'transl_z']] = pcd_center
building_gdf[['pts_number']] = len(segment.points)
print(building_gdf.head(1))
# 8. 2D to 3D
# Generate the vertice list
vertices = list(building_vector.exterior.coords)
# Construct the Open3D Object
polygon_2d = o3d.geometry.LineSet()
polygon_2d.points = o3d.utility.Vector3dVector([point + (0,) for point in vertices])
polygon_2d.lines = o3d.utility.Vector2iVector([[i, (i + 1) % len(vertices)] for i in range(len(vertices))])
# o3d. visualization.draw_geometries([polygon_2d])
# 8.2 The top layer
extrusion = o3d.geometry.LineSet()
extrusion.points = o3d.utility.Vector3dVector([point + (height,) for point in vertices])
extrusion.lines = o3d.utility.Vector2iVector([[i, (i + 1) % len(vertices)] for i in range(len(vertices))])
# o3d. visualization.draw_geometries([polygon_2d, extrusion])
# Plot the vertices
temp = polygon_2d + extrusion
temp.points
temp_o3d = o3d.geometry.PointCloud()
temp_o3d.points = temp.points
# o3d.visualization.draw_geometries([temp_o3d])
# 9. Generating the base vertices for the 3D mesh
a = np.array(building_vector.exterior.coords)
b = np.ones([a.shape[0],1])*sample.get_center()[2]
c = np.ones([a.shape[0],1])*(sample.get_center()[2] + height)
ground_pc = np.hstack((a,b))
up_pc = np.hstack((a,c))
temp_o3d = o3d.geometry.PointCloud()
temp_o3d.points = o3d.utility.Vector3dVector(np.concatenate((ground_pc, up_pc), axis=0))
# o3d.visualization.draw_geometries([temp_o3d])
# 10. Computing the alpha shapre of the 3D base points
alpha = 20
mesh = o3d.geometry.TriangleMesh.create_from_point_cloud_alpha_shape(temp_o3d, alpha)
mesh.compute_vertex_normals()
mesh.paint_uniform_color([0.5, 0.4, 0])
# o3d.visualization.draw_geometries([temp_o3d, mesh, segment], mesh_show_back_face=True)
# 11. Post-Porcessing Operations and Export
mesh.translate(pcd_center)
o3d.io.write_triangle_mesh("building_mesh.ply", mesh, write_ascii=True, compressed=True, write_vertex_normals=False, write_vertex_colors=False, write_triangle_uvs=False)
building_gdf.to_file("building_gdf.shp")