-
Notifications
You must be signed in to change notification settings - Fork 11
/
utils.py
205 lines (170 loc) · 7.41 KB
/
utils.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
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
import numpy as np
import open3d as o3d
import cv2
from scipy.spatial.transform import Rotation
from SIFT import *
def transform_3d_points(points_3d, transformation):
"""
:param pcd:
:param transformation:
:return: Transformed pcd
"""
source_pcd = o3d.geometry.PointCloud()
source_pcd.points = o3d.utility.Vector3dVector(points_3d)
source_pcd.transform(transformation)
return np.asarray(source_pcd.points)
def geometrically_averaged_points(points, points_unique, points_index):
"""
Return 3D Geometrically averaged points
:param points:
:param points_unique:
:return:
"""
avg_points = np.zeros(shape=points_unique.shape)
for i in range(len(points_unique)):
temp = []
# count = 0
for pt_idx in range(len(points)):
if points_index[pt_idx] == i:
temp.append(points[pt_idx])
temp = np.array(temp)
avg_points[i] = np.mean(temp, axis=0, dtype=np.float64)
return avg_points
def preprocess_point_cloud(pcd, voxel_size):
'''
:param pcd: Point cloud dataset
:param voxel_size: Voxel size of dataset
:return: downsampled point cloud
'''
pcd_down = pcd.voxel_down_sample(voxel_size)
radius_normal = voxel_size * 2
pcd_down.estimate_normals(o3d.geometry.KDTreeSearchParamHybrid(radius=radius_normal, max_nn=30))
radius_feature = voxel_size * 5
pcd_fpfh = o3d.pipelines.registration.compute_fpfh_feature(pcd_down, o3d.geometry.KDTreeSearchParamHybrid(radius=radius_feature, max_nn=100))
return pcd_down, pcd_fpfh
def get_boundary(source_pcd):
# Intrinsic parameter for Realsense D415
depth_scaling_factor = 999.99
focal_length = 597.522
img_center_x = 312.885
img_center_y = 239.870
x_min = np.min(np.asarray(source_pcd.points)[:, 0])
x_max = np.max(np.asarray(source_pcd.points)[:, 0])
y_min = np.min(np.asarray(source_pcd.points)[:, 1])
y_max = np.max(np.asarray(source_pcd.points)[:, 1])
x_min_idx = np.where(np.asarray(source_pcd.points)[:, 0] == x_min)
x_max_idx = np.where(np.asarray(source_pcd.points)[:, 0] == x_max)
y_min_idx = np.where(np.asarray(source_pcd.points)[:, 1] == y_min)
y_max_idx = np.where(np.asarray(source_pcd.points)[:, 1] == y_max)
u_min = x_min * focal_length / (np.asarray(source_pcd.points)[x_min_idx][0][2]) + img_center_x
u_max = x_max * focal_length / (np.asarray(source_pcd.points)[x_max_idx][0][2]) + img_center_x
v_min = y_min * focal_length / (np.asarray(source_pcd.points)[y_min_idx][0][2]) + img_center_y
v_max = y_max * focal_length / (np.asarray(source_pcd.points)[y_max_idx][0][2]) + img_center_y
return u_min, u_max, v_min, v_max
def R_t_matrix_to_vector(R_t):
print('R|t shape:', np.array(R_t)[:3][:, :3])
r = Rotation.from_matrix(np.array(R_t)[:3][:, :3]) # Rodriguess
r = r.as_quat()
qx = r[0]
qy = r[1]
qz = r[2]
qw = r[3]
tx = R_t[0,3]
ty = R_t[0, 3]
tz = R_t[0, 3]
rotation = np.array([qx, qy, qz])
translation = np.array([tx, ty, tz])
return rotation, translation
def vector_to_matrix(vector):
transformation_matrix = np.identity(4)
rotation_matrix = np.identity(3)
rvecs = vector[:3]
cv2.Rodrigues(rvecs, rotation_matrix)
transformation_matrix[0:3, 0:3] = rotation_matrix
transformation_matrix[0][3] = vector[3]
transformation_matrix[1][3] = vector[4]
transformation_matrix[2][3] = vector[5]
print('transformation_matrix:', transformation_matrix)
return transformation_matrix
def get_cam_indices(pts1, pts2, pts3):
cam0 = np.array([0] * len(pts1))
cam1 = np.array([1] * len(pts2))
cam2 = np.array([2] * len(pts3))
# cam3 = np.array([3] * len(pts4))
result = np.concatenate([cam0, cam1, cam2])
print('cam indices:', len(result))
return result
def get_point_indices(pts1_3d, pts2_3d, pts3_3d, pts_unique):
result = []
for i in range(len(pts1_3d)):
for j in range(len(pts_unique)):
if pts1_3d[i][0] == pts_unique[j][0] and pts1_3d[i][1] == pts_unique[j][1] and pts1_3d[i][2] == pts_unique[j][2]:
result.append(j)
break
for i in range(len(pts2_3d)):
for j in range(len(pts_unique)):
if pts2_3d[i][0] == pts_unique[j][0] and pts2_3d[i][1] == pts_unique[j][1] and pts2_3d[i][2] == pts_unique[j][2]:
result.append(j)
break
for i in range(len(pts3_3d)):
for j in range(len(pts_unique)):
if pts3_3d[i][0] == pts_unique[j][0] and pts3_3d[i][1] == pts_unique[j][1] and pts3_3d[i][2] == pts_unique[j][2]:
result.append(j)
break
# for i in range(len(pts4_3d)):
# for j in range(len(pts_unique)):
# if pts4_3d[i][0] == pts_unique[j][0] and pts4_3d[i][1] == pts_unique[j][1] and pts4_3d[i][2] == pts_unique[j][2]:
# result.append(j)
# break
print('point indices:', len(result))
return np.asarray(result)
def full_registration(pcds, max_correspondence_distance_coarse,
max_correspondence_distance_fine,
transformation_icp,
information_icp):
pose_graph = o3d.pipelines.registration.PoseGraph()
odometry = np.identity(4)
pose_graph.nodes.append(o3d.pipelines.registration.PoseGraphNode(odometry))
n_pcds = len(pcds)
for source_id in range(n_pcds):
for target_id in range(source_id + 1, n_pcds):
print("Build o3d.pipelines.registration.PoseGraph")
if target_id == source_id + 1: # odometry case
odometry = np.dot(transformation_icp, odometry)
pose_graph.nodes.append(
o3d.pipelines.registration.PoseGraphNode(
np.linalg.inv(odometry)))
pose_graph.edges.append(
o3d.pipelines.registration.PoseGraphEdge(source_id,
target_id,
transformation_icp,
information_icp,
uncertain=False))
else: # loop closure case
pose_graph.edges.append(
o3d.pipelines.registration.PoseGraphEdge(source_id,
target_id,
transformation_icp,
information_icp,
uncertain=True))
return pose_graph
def reproject_point2d(points, depth_source_path):
"""
:param points:
:return: 3d points
"""
depth_scaling_factor = 999.99
focal_length = 597.522 ## mm
img_center_x = 312.885
img_center_y = 239.870
depth = np.array(o3d.io.read_image(depth_source_path), np.float32)
points_3d = np.array([])
for point in points:
u = point[0]
v = point[1]
# Normalized image plane -> (u, v, 1) * z = zu, zv, z
z = np.asarray(depth, dtype=np.float64)[np.int32(v)][np.int32(u)] / depth_scaling_factor # in mm distance
x = (u - img_center_x) * z / focal_length
y = (v - img_center_y) * z / focal_length
points_3d = np.append(points_3d, np.array([x, y, z], dtype=np.float32)).reshape(-1, 3)
return points_3d