-
Notifications
You must be signed in to change notification settings - Fork 3
/
StereoChess.py
368 lines (292 loc) · 16.1 KB
/
StereoChess.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
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
''' CONFIDENTIAL
Copyright (c) 2021 Eugeniu Vezeteu,
Department of Remote Sensing and Photogrammetry,
Finnish Geospatial Research Institute (FGI), National Land Survey of Finland (NLS)
PERMISSION IS HEREBY LIMITED TO FGI'S INTERNAL USE ONLY. THE CODE
MAY BE RE-LICENSED, SHARED, OR TAKEN INTO OTHER USE ONLY WITH
A WRITTEN CONSENT FROM THE HEAD OF THE DEPARTMENT.
The software is provided "as is", without warranty of any kind, express or
implied, including but not limited to the warranties of merchantability,
fitness for a particular purpose and noninfringement. In no event shall the
authors or copyright holders be liable for any claim, damages or other
liability, whether in an action of contract, tort or otherwise, arising from,
out of or in connection with the software or the use or other dealings in the
software.
'''
import numpy as np
import cv2
import glob
import pickle
from mpl_toolkits.mplot3d import Axes3D
import matplotlib.pyplot as plt
import pandas as pd
from utils import *
np.set_printoptions(suppress=True)
class StereoChess_Calibrator(object):
def __init__(self, filepath, name = ''):
self.term_criteria=(cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_COUNT, 1000, 0.0001)
self.square = 0.1 #m
self.objp = np.zeros((10 * 7, 3), np.float32)
self.objp[:, :2] = np.mgrid[0:10, 0:7].T.reshape(-1, 2) * self.square
self.objpoints = [] # 3d point in real world space
self.imgpoints_l = [] # 2d points in image plane.
self.imgpoints_r = [] # 2d points in image plane.
self.cal_path = filepath
self.image_width = 1936
self.image_height = 1216
self.image_size = None
self.fx = 0.25
self.fy = 0.3
self.position = (20, 30)
self.see = True
self.flipVertically = False
self.name = name
self.total = 0
def readMonoData(self):
left_outside = load_obj(name='chess_{}_left'.format(self.name))
self.K_left = left_outside['K']
self.D_left = left_outside['D']
right_outside = load_obj(name='chess_{}_right'.format(self.name))
self.K_right = right_outside['K']
self.D_right = right_outside['D']
print("Left K:\n", self.K_left)
print("Left distortion : ", self.D_left.ravel())
print("Right K:\n", self.K_right)
print("Right distortion : ", self.D_right.ravel())
print('Loaded mono internal calibration data')
def readStereoData(self):
camera_model = load_obj('{}_chess_camera_model'.format(self.name ))
camera_model_rectify = load_obj('{}_chess_camera_model_rectify'.format(self.name))
self.K_left = camera_model['K_left']
self.K_right = camera_model['K_right']
self.D_left = camera_model['D_left']
self.D_right = camera_model['D_right']
self.R = camera_model['R']
self.T = camera_model['T']
self.Q = camera_model_rectify['Q']
self.roi_left, self.roi_right = camera_model_rectify['roi_left'],camera_model_rectify['roi_right']
self.leftMapX, self.leftMapY = camera_model_rectify['leftMapX'], camera_model_rectify['leftMapY']
self.rightMapX, self.rightMapY = camera_model_rectify['rightMapX'], camera_model_rectify['rightMapY']
print('Rotation R')
print(self.R)
print('Translation T')
print(self.T)
print('Stereo data has been loaded...')
def read_images(self, test=False):
images_right = glob.glob(self.cal_path + '/Right/*.png')
images_left = glob.glob(self.cal_path + '/Left/*.png')
images_left.sort()
images_right.sort()
self.LeftImg, self.RightImg = [], []
wait = 0
for i, fname in enumerate(images_right):
img_l = cv2.imread(images_left[i])
img_r = cv2.imread(images_right[i])
if self.flipVertically:
img_l = cv2.flip(img_l, -1)
img_r = cv2.flip(img_r, -1)
self.LeftImg.append(img_l)
self.RightImg.append(img_r)
gray_l = cv2.cvtColor(img_l, cv2.COLOR_BGR2GRAY)
gray_r = cv2.cvtColor(img_r, cv2.COLOR_BGR2GRAY)
if test:
ret_l = True
ret_r = True
else:
ret_l, corners_l = cv2.findChessboardCorners(gray_l, (10, 7), None)
ret_r, corners_r = cv2.findChessboardCorners(gray_r, (10, 7), None)
if ret_l and ret_r:
if test == False:
self.objpoints.append(self.objp)
rt = cv2.cornerSubPix(gray_l, corners_l, (11, 11), (-1, -1), self.term_criteria)
self.imgpoints_l.append(corners_l)
ret_l = cv2.drawChessboardCorners(img_l, (10, 7), corners_l, ret_l)
rt = cv2.cornerSubPix(gray_r, corners_r, (11, 11), (-1, -1), self.term_criteria)
self.imgpoints_r.append(corners_r)
ret_r = cv2.drawChessboardCorners(img_r, (10, 7), corners_r, ret_r)
self.total+=len(corners_r)
if self.see:
cam_right_resized = cv2.resize(img_r, None, fx=self.fx, fy=self.fy)
cam_left_resized = cv2.resize(img_l, None, fx=self.fx, fy=self.fy)
cv2.putText(cam_left_resized, "Left", self.position, cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255, 0))
cv2.putText(cam_right_resized, "Right", self.position, cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255, 0))
im_h = cv2.hconcat([cam_left_resized, cam_right_resized])
cv2.imshow('Stereo camera', im_h)
k = cv2.waitKey(wait)
if k % 256 == 32: # pressed space
self.see = False
cv2.destroyAllWindows()
elif k & 0xFF == ord('q'):
wait = 50
self.img_shape = gray_l.shape[::-1]
cv2.destroyAllWindows()
def stereoCalibrate(self, flags=None, save=False):
if flags is None:
flags = 0
flags |= cv2.CALIB_USE_INTRINSIC_GUESS
flags |= cv2.CALIB_FIX_FOCAL_LENGTH
flags |= cv2.CALIB_FIX_ASPECT_RATIO
self.rms_stereo, self.K_left, self.D_left, self.K_right, self.D_right, self.R, self.T, self.E, self.F = cv2.stereoCalibrate(
self.objpoints, self.imgpoints_l,
self.imgpoints_r, self.K_left, self.D_left, self.K_right, self.D_right, self.img_shape,
criteria=self.term_criteria, flags=flags)
print('Stereo calibraion done')
print('rms_stereo:{}'.format(self.rms_stereo))
print('Rotation R')
print(self.R)
print('Translation T')
print(self.T)
R1, R2, P1, P2, self.Q, self.roi_left, self.roi_right = cv2.stereoRectify(self.K_left, self.D_left, self.K_right,
self.D_right, self.img_shape, self.R, self.T,
flags=cv2.CALIB_ZERO_DISPARITY, alpha=-1)
self.leftMapX, self.leftMapY = cv2.initUndistortRectifyMap(
self.K_left, self.D_left, R1,
P1, self.img_shape, cv2.CV_32FC1)
self.rightMapX, self.rightMapY = cv2.initUndistortRectifyMap(
self.K_right, self.D_right, R2,
P2, self.img_shape, cv2.CV_32FC1)
if save:
camera_model_rectify = dict([('R1', R1), ('R2', R2), ('P1', P1),
('P2', P2), ('Q', self.Q),
('roi_left', self.roi_left), ('roi_right', self.roi_right),
('leftMapX', self.leftMapX), ('leftMapY', self.leftMapY),
('rightMapX', self.rightMapX), ('rightMapY', self.rightMapY)])
camera_model = dict([('K_left', self.K_left), ('K_right', self.K_right), ('D_left', self.D_left),
('D_right', self.D_right), ('R', self.R), ('T', self.T),
('E', self.E), ('F', self.F)])
save_obj(camera_model,self.name+'_chess_camera_model')
save_obj(camera_model_rectify, self.name+'_chess_camera_model_rectify')
def depth_map_SGBM(self, imgL, imgR):
window_size = 5
if self.left_matcher is None:
self.left_matcher = cv2.StereoSGBM_create(minDisparity=5,
numDisparities=256,
blockSize=window_size,
)
self.stereo = cv2.StereoBM_create(numDisparities=256, blockSize=5)
self.right_matcher = cv2.ximgproc.createRightMatcher(self.left_matcher)
# FILTER Parameters
lmbda = 80000
sigma = 1.3
self.wls_filter = cv2.ximgproc.createDisparityWLSFilter(matcher_left=self.left_matcher)
self.wls_filter.setLambda(lmbda)
self.wls_filter.setSigmaColor(sigma)
displ = self.left_matcher.compute(imgL, imgR) # .astype(np.float32)/16
dispr = self.right_matcher.compute(imgR, imgL) # .astype(np.float32)/16
displ = np.int16(displ)
dispr = np.int16(dispr)
SGBM_disp = self.wls_filter.filter(displ, imgL, imgR, dispr)
SGBM_disp = cv2.normalize(src=SGBM_disp, dst=SGBM_disp, beta=0, alpha=255, norm_type=cv2.NORM_MINMAX)
SGBM_disp = np.uint8(SGBM_disp)
disparity = self.stereo.compute(imgL, imgR)
disparity = cv2.normalize(src=disparity, dst=disparity, beta=0, alpha=255, norm_type=cv2.NORM_MINMAX)
BM_disp = np.uint8(disparity)
return SGBM_disp, BM_disp
def depth_and_color(self, img, left_rectified):
points = cv2.reprojectImageTo3D(img, self.Q)
reflect_matrix = np.identity(3) # reflect on x axis
reflect_matrix[0] *= -1
points = np.matmul(points, reflect_matrix)
colors = cv2.cvtColor(left_rectified, cv2.COLOR_BGR2RGB)
# filter by min disparity
mask = img > img.min()
out_points = points[mask]
out_colors = colors[mask]
idx = np.fabs(out_points[:, 0]) < 250 # 10.5 # filter by dimension
out_points = out_points[idx]
out_colors = out_colors.reshape(-1, 3)
out_colors = out_colors[idx]
write_ply('BM_2.ply', out_points, out_colors)
print('%s saved' % 'out.ply')
reflected_pts = np.matmul(out_points, reflect_matrix)
projected_img, _ = cv2.projectPoints(reflected_pts, np.identity(3), np.array([0., 0., 0.]), self.K_left,
self.D_left)
projected_img = projected_img.reshape(-1, 2)
blank_img = np.zeros(left_rectified.shape, 'uint8')
img_colors = left_rectified[mask][idx].reshape(-1, 3)
for i, pt in enumerate(projected_img):
pt_x = int(pt[0])
pt_y = int(pt[1])
if pt_x > 0 and pt_y > 0:
# use the BGR format to match the original image type
col = (int(img_colors[i, 2]), int(img_colors[i, 1]), int(img_colors[i, 0]))
cv2.circle(blank_img, (pt_x, pt_y), 1, col)
return blank_img
def depth(self):
self.left_matcher = None
wait = 0
for i, fname in enumerate(self.LeftImg):
leftFrame = self.LeftImg[i]
rightFrame = self.RightImg[i]
left_rectified = cv2.remap(leftFrame, self.leftMapX, self.leftMapY, cv2.INTER_LINEAR, cv2.BORDER_CONSTANT)
right_rectified = cv2.remap(rightFrame, self.rightMapX, self.rightMapY, cv2.INTER_LINEAR,
cv2.BORDER_CONSTANT)
#out = right_rectified.copy()
#out[:, :, 0] = left_rectified[:, :, 0]
#out[:, :, 1] = left_rectified[:, :, 1]
#out[:, :, 2] = right_rectified[:, :, 2]
gray_left = cv2.cvtColor(left_rectified, cv2.COLOR_BGR2GRAY)
gray_right = cv2.cvtColor(right_rectified, cv2.COLOR_BGR2GRAY)
#disparity_image2 = cv2.applyColorMap(disparity_image2, cv2.COLORMAP_JET)
SGBM_disp, BM_disp = self.depth_map_SGBM(gray_left, gray_right, ) # Get the disparity map
img_top = cv2.hconcat(
[cv2.resize(left_rectified, None, fx=self.fx, fy=self.fy), cv2.resize(right_rectified, None, fx=self.fx, fy=self.fy)])
img_bot = np.hstack((cv2.resize(SGBM_disp, None, fx=self.fx, fy=self.fy), cv2.resize(BM_disp, None, fx=self.fx, fy=self.fy)))
#pointCloudColor = self.depth_and_color(img=BM_disp.copy(),left_rectified=left_rectified)
pointCloudColor = self.depth_and_color(img=SGBM_disp.copy(),left_rectified=left_rectified)
cv2.imshow('img_bot ', img_bot)
cv2.imshow('Result', img_top)
cv2.imshow('pointCloudColor', cv2.resize(pointCloudColor, None, fx=.4, fy=.4))
k = cv2.waitKey(wait)
if k & 0xFF == ord('q'):
wait = 1
#break
cv2.waitKey(0)
cv2.destroyAllWindows()
def calibrationReport(self):
Distorsion_models = {'ST': ['Standard', 0, 'Standard'],
'RAT': ['Rational', cv2.CALIB_RATIONAL_MODEL, 'CALIB_RATIONAL_MODEL'],
'THP': ['Thin Prism', cv2.CALIB_THIN_PRISM_MODEL, 'CALIB_THIN_PRISM_MODEL'],
'TIL': ['Tilded', cv2.CALIB_TILTED_MODEL, 'CALIB_TILTED_MODEL'], # }
'RAT+THP': ['Rational+Thin Prism',
cv2.CALIB_RATIONAL_MODEL + cv2.CALIB_THIN_PRISM_MODEL,
'CALIB_RATIONAL_MODEL + CALIB_THIN_PRISM_MODEL'],
'THP+TIL': ['Thin Prism+Tilded', cv2.CALIB_THIN_PRISM_MODEL + cv2.CALIB_TILTED_MODEL,
'CALIB_THIN_PRISM_MODEL + CALIB_TILTED_MODEL'],
'RAT+TIL': ['Rational+Tilded', cv2.CALIB_RATIONAL_MODEL + cv2.CALIB_TILTED_MODEL,
'CALIB_RATIONAL_MODEL + CALIB_TILTED_MODEL'],
'CMP': ['Complete',
cv2.CALIB_RATIONAL_MODEL + cv2.CALIB_THIN_PRISM_MODEL + cv2.CALIB_TILTED_MODEL,
'Complete']}
def rot2eul(R):
beta = -np.arcsin(R[2, 0])
alpha = np.arctan2(R[2, 1] / np.cos(beta), R[2, 2] / np.cos(beta))
gamma = np.arctan2(R[1, 0] / np.cos(beta), R[0, 0] / np.cos(beta))
alpha, beta, gamma = math.degrees(alpha), math.degrees(beta), math.degrees(gamma)
return np.array((alpha, beta, gamma))
calibration_results = pd.DataFrame(
{"params": [ 'Tx','Ty','Tz','Rx','Ry','Rz', 'Error']})
rms_all = ['Error']
min_error, flag = 10000000, 0
for key in Distorsion_models:
print()
print(key, '->', Distorsion_models[key][0], ' , ', Distorsion_models[key][1], ' , ',
Distorsion_models[key][2])
flags = Distorsion_models[key][1]
self.stereoCalibrate(flags=flags, save=False)
s = np.append(self.T, rot2eul(self.R))
calibration_results[str(key)] = pd.Series(s)
calibration_results.fillna('---', inplace=True)
rms_all.append(self.rms_stereo)
calibration_results[str(key)] = calibration_results[str(key)].map(
lambda x: round(x, 4) if isinstance(x, (int, float)) else x).astype(str)
if self.rms_stereo < min_error:
min_error=self.rms_stereo
flag = flags
calibration_results.iloc[-1, :] = rms_all
calibration_results.iloc[-1, :] = calibration_results.iloc[-1, :].map(
lambda x: round(x, 4) if isinstance(x, (int, float)) else x)
save_csv(obj=calibration_results, name=self.name+'_StereoCalibration')
print('Best distortion model for stereo calib is {}'.format(flag))
self.stereoCalibrate(flags=flag,save=True)
return flag