-
Notifications
You must be signed in to change notification settings - Fork 0
/
ICP.py
218 lines (160 loc) · 8.03 KB
/
ICP.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
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
import cv2
import numpy as np
import matplotlib.pyplot as plt
import plotting
import util
from scipy.spatial.distance import directed_hausdorff
import torch
import os
from sklearn.metrics.pairwise import cosine_similarity
import matplotlib.pyplot as plt
import math
import knn
import random
from scipy.spatial.distance import cdist
#from simpleicp import PointCloud, SimpleICP
import Icp2d
import multiview_utils
def icp(data_ref, data_sync, best_shift_array, best_scale_array, sync_dict_array, save_dir = None, name = '_cam_'):
icp_rot_array = []
icp_init_rot_array = []
icp_shift_array = []
init_ref_center_array = []
init_sync_center_array = []
time_shift_array = []
time_scale_array = []
#sync_dict_array = []
index_array = []
for i in range(len(data_sync)):
best_shift = best_shift_array[i]
best_scale = best_scale_array[i]
sync_dict = sync_dict_array[i]
init_d, init_ref_center, init_sync_center, reflect_true = rot_shift_search_time_weight(data_ref, data_sync[i], sync_dict, best_scale, best_shift, save_dir = save_dir, name = str(i) + name)
icp_rot, icp_shift, icp_init_rot, index = Icp2d.icp_normalize(data_ref, data_sync[i], sync_dict, init_d, init_ref_center, init_sync_center, reflect_true, save_dir = save_dir, name = str(i), best_scale = best_scale)
icp_rot_array.append(icp_rot)
icp_init_rot_array.append(icp_init_rot)
icp_shift_array.append(icp_shift)
init_ref_center_array.append(init_ref_center)
init_sync_center_array.append(init_sync_center)
time_shift_array.append(best_shift)
time_scale_array.append(best_scale)
sync_dict_array.append(sync_dict)
return icp_rot_array, icp_init_rot_array, icp_shift_array, init_ref_center_array, init_sync_center_array, time_shift_array, time_scale_array, sync_dict_array, index_array
def rot_shift_search_time_weight(data_ref, data_sync, sync_time_dict, scale, offset, save_dir = None, name = None):
if os.path.isdir(save_dir + '/search_rot') == False:
os.mkdir(save_dir + '/search_rot')
if os.path.isdir(save_dir + '/search_rot/' + name) == False:
os.mkdir(save_dir + '/search_rot/' + name)
ref_coords = []
ref_time = []
ref_cor_time = []
ref_dict = {}
#print(sync_time_dict, " sync dict")
#print("**********************************************")
for fr in list(sync_time_dict.keys()):
ref_dict[fr] = []
for tr in list(data_ref[fr].keys()):
ref_coords.append(data_ref[fr][tr][0:2])
ref_time.append(fr)
ref_cor_time.append(scale*sync_time_dict[fr] + offset)
ref_dict[fr].append(data_ref[fr][tr][0:2])
#print(scale, offset, " scale and offsetttt")
sync_coords = []
sync_time = []
sync_cor_time = []
sync_dict = {}
for item in sync_time_dict.items():
sync_dict[item[1]] = []
for tr in list(data_sync[item[1]].keys()):
sync_coords.append(data_sync[item[1]][tr][0:2])
sync_time.append(item[1])
sync_cor_time.append(item[0])
sync_dict[item[1]].append(data_sync[item[1]][tr][0:2])
#print(sync_cor_time)
ref_center = np.mean(ref_coords, axis = 0)
sync_center = np.mean(sync_coords, axis = 0)
best_d = 0
best_error = np.inf
best_x = 1
best_y = 1
all_angles = []
all_error = []
reflect_true = False
#################################################
normalize_ref = np.array(ref_coords) - ref_center
normalize_rot = np.array(sync_coords) - sync_center#ref_center
ref_x_size = 1.0#max([abs(normalize_ref_x_max), abs(normalize_ref_x_min)])
ref_y_size = 1.0#max([abs(normalize_ref_y_max), abs(normalize_ref_y_min)])
rot_x_size = 1.0#max([abs(normalize_rot_x_max), abs(normalize_rot_x_min)])
rot_y_size = 1.0#max([abs(normalize_rot_y_max), abs(normalize_rot_y_min)])
ref_normalize = np.transpose(np.stack([normalize_ref[:, 0]/ref_x_size, normalize_ref[:, 1]/ref_y_size]))
sync_normalize = np.transpose(np.stack([normalize_rot[:, 0]/rot_x_size, normalize_rot[:, 1]/rot_y_size]))
sync_space_time = []
ref_space_time = []
sync_time = []
ref_time = []
for i in sync_time_dict.keys():
sync_time.append((i*np.ones(np.array(sync_dict[sync_time_dict[i]])[:, 1].shape)))
ref_time.append((i*np.ones(np.array(ref_dict[i])[:, 1].shape)))
sync_normalize1 = np.transpose(np.stack([(np.array(sync_dict[sync_time_dict[i]])[:, 0] - sync_center[0])/rot_x_size, (np.array(sync_dict[sync_time_dict[i]])[:, 1] - sync_center[1])/rot_y_size]))
ref_normalize1 = np.transpose(np.stack([(np.array(ref_dict[i])[:, 0] - ref_center[0])/ref_x_size, (np.array(ref_dict[i])[:, 1] - ref_center[1])/ref_y_size]))
sync_space_time.append(sync_normalize1)
ref_space_time.append(ref_normalize1)
sync_normalize1 = np.concatenate(sync_space_time)
ref_normalize1 = np.concatenate(ref_space_time)
sync_time = np.concatenate(sync_time)
ref_time = np.concatenate(ref_time)
for reflect in [False]:
for d in range(50,300):
rot_coords1, R, o, p, s = rotate(sync_normalize1, origin=[0,0], shift=[0,0], degrees=d, reflect = reflect)
ref_normalize1_time = np.transpose(np.stack([ref_normalize1[:, 0], ref_normalize1[:, 1], ref_time]))
rot_coords1_time = np.transpose(np.stack([rot_coords1[:, 0], rot_coords1[:, 1], sync_time]))
error = multiview_utils.chamfer_distance(ref_normalize1_time, rot_coords1_time, metric='l2', direction='bi')
if best_error > error:
best_d = d
best_error = error
reflect_true = reflect
all_angles.append(d)
all_error.append(error)
rot_normalize, R, o, p, s = rotate(sync_normalize1, origin=[0,0], shift=[0,0], degrees=best_d, reflect = reflect_true, scale_x = best_x, scale_y = best_y)
error = best_error
if save_dir is not None:
fig, ax1 = plt.subplots(1, 1)
ax1.set_yscale("linear")
ax1.scatter(all_angles, all_error)
if name is not None:
fig.savefig(save_dir + '/search_rot/' + name + '/error_' + str(name) + '_' + str(best_d) + '_degree_' + '.png')
else:
fig.savefig(save_dir + '/search_rot/' + name + '/' + 'error_' + str(best_d) + '_degree_' + '.png')
plt.close('all')
if save_dir is not None:
fig, ax1 = plt.subplots(1, 1)
ax1.scatter(np.array(rot_normalize)[:, 0], np.array(rot_normalize)[:, 1], c = 'r')
ax1.scatter(np.array(ref_normalize)[:, 0], np.array(ref_normalize)[:, 1], c = 'b')
ax1.set_title(str(error))
fig1, ax2 = plt.subplots(1, 1)
ax2.scatter(np.array(sync_normalize)[:, 0], np.array(sync_normalize)[:, 1], c = 'r')
ax2.scatter(np.array(ref_normalize)[:, 0], np.array(ref_normalize)[:, 1], c = 'b')
ax2.set_title("init")
if name is not None:
fig.savefig(save_dir + '/search_rot/' + name + '/best_' + str(name) + '_' + str(best_d) + '_degree_' + '.png')
else:
fig.savefig(save_dir + '/search_rot/' + name + '/' + 'best_' + str(best_d) + '_degree_' + '.png')
fig1.savefig(save_dir + '/search_rot/' + name + '/init.png')
plt.close('all')
return math.radians(best_d), ref_center, sync_center, reflect_true
def rotate(p, origin=(0, 0), shift=(0,0), degrees=0.0, reflect = False, scale_x = 1.0, scale_y = 1.0):
angle = np.deg2rad(degrees)
R = np.array([[scale_x*np.cos(angle), -np.sin(angle)],
[np.sin(angle), scale_y*np.cos(angle)]])
if reflect == True:
R = np.array([[scale_x*np.cos(angle), -np.sin(angle)],
[np.sin(angle), -1*scale_y*np.cos(angle)]])
o = np.atleast_2d(origin)
p = np.atleast_2d(p)
s = np.atleast_2d(shift)
#print(p.shape, o.shape, " HELLO")
if p.shape[0] > 1:
return np.squeeze((R @ (p.T-o.T) + s.T).T), R, o, p, s
else:
return (R @ (p.T-o.T) + s.T).T, R, o, p, s