camera-localization/calib.py at master · orrblue/camera-localization · GitHub

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
'''
calib.py

Attempt at Camera Calibration using homography/matching
AKA, no checker board!

So far, no successful calibration. So far...


'''

import numpy as np
import cv2
import matplotlib.pyplot as plt
import glob


pattern = cv2.imread('img/pattern.png')
pattern = cv2.cvtColor(pattern, cv2.COLOR_BGR2GRAY)

images = glob.glob('img/*.JPG')
pattern_pts = []
scene_pts = []
best_matches_size = 10

for img_name in images:
    img = cv2.imread(img_name)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    orb = cv2.ORB_create()

    # detect features in plain QR code, then in provided images of QR code
    keypoints_pattern, descriptors_pattern = orb.detectAndCompute(pattern, None)
    keypoints_scene, descriptors_scene = orb.detectAndCompute(gray, None)
    bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)

    # match features and sort based on quality. Pick out n-th best matches.
    matches = bf.match(descriptors_pattern, descriptors_scene)
    matches = sorted(matches, key = lambda x: x.distance)
    best_matches = matches[ : best_matches_size]

    #add n-th best matches to lists of object pts, and image pts
    for i in range(best_matches_size):
        pattern_pts.append(keypoints_pattern[ best_matches[i].queryIdx ].pt)
        scene_pts.append(keypoints_scene[ best_matches[i].trainIdx ].pt)


# reshape to # num_elms x best_matches_size x 2
pattern_pts = np.float32(pattern_pts).reshape(-1, best_matches_size, 2)
scene_pts = np.float32(scene_pts).reshape(-1, best_matches_size, 2)

# insert 0 as Z-coordinate since object is a plane
rows = pattern_pts.shape[0]
pattern_pts = np.insert(pattern_pts, 2, np.zeros((rows, 1)), 2)


# Source cam_mat for calibCamera() to use as estimate. Focal length in pixels
camera_matrix_source = np.array([
    [2768, 0., 1224.],
    [0., 2770, 1632.],
    [0., 0., 1.]])

# camera matrix, distortion coefficients, rotation vector, translation
rms, camera_matrix, dist_coefs, rvecs, tvecs = cv2.calibrateCamera(pattern_pts, scene_pts,
    (2448, 3264), cameraMatrix = camera_matrix_source, distCoeffs = None, rvecs = None,
    tvecs = None, flags=1)

print()
print(rms)
print()
print(camera_matrix)
print()
print(dist_coefs)


# showing undistorted image
match1 = cv2.imread('img/IMG_6725.JPG', 0)
undist1 = cv2.undistort(match1, camera_matrix, dist_coefs, None)
output = cv2.resize(undist1, None, fx=0.3, fy=0.3,interpolation = cv2.INTER_CUBIC)
cv2.imshow('output', output)

cv2.waitKey(0)
cv2.destroyAllWindows()


'''

=== TRIALS ===

best_matches_size = 10
RMS = 54.50464398008055

Camera Matrix =
[[  1.50512671e+02   0.00000000e+00   1.08574478e+03]
 [  0.00000000e+00   9.14708563e+01   1.66977834e+03]
 [  0.00000000e+00   0.00000000e+00   1.00000000e+00]]

Distortion coefficients =
[[ 0.59014476 -0.0191317   0.09039141 -0.31884398 -0.00223853]]


12
66.53174503437471

[[  1.83370933e+02   0.00000000e+00   1.16538885e+03]
 [  0.00000000e+00   1.34968339e+02   1.63702049e+03]
 [  0.00000000e+00   0.00000000e+00   1.00000000e+00]]

[[-0.05248756  0.00591458 -0.01091695 -0.04493884 -0.00028817]]


10 + flags = 1 'CV_CALIB_USE_INTRINSIC_GUESS'
78.64751038656593
source focal length was incorrectly in mm --> 4.15mm

[[  4.15000000e+00   0.00000000e+00   1.22400000e+03]
 [  0.00000000e+00   4.15000000e+00   1.63200000e+03]
 [  0.00000000e+00   0.00000000e+00   1.00000000e+00]]

[[  4.61762188e-13   1.37063213e-09  -1.37654984e-14  -8.85760712e-15
   -2.11943788e-13]]


10 + flags = 1 (correct focal length this time)
57.00925747352173

[[  3.19058750e+03   0.00000000e+00   1.33308949e+03]
 [  0.00000000e+00   1.18929582e+03   1.68754358e+03]
 [  0.00000000e+00   0.00000000e+00   1.00000000e+00]]

[[  1.27001459e+01   4.19172651e+02  -5.27023604e-01   1.63341525e+00
   -7.79681554e+03]]


12 + flags = 2
67.58625900752637

[[  1.64211532e+02   0.00000000e+00   1.13830361e+03]
 [  0.00000000e+00   1.64330181e+02   1.63379828e+03]
 [  0.00000000e+00   0.00000000e+00   1.00000000e+00]]

[[-0.08502798  0.01974622 -0.02661665 -0.03478743 -0.0013597 ]]


10 + flags = 2
54.969109577357244

[[  1.89269215e+01   0.00000000e+00   1.11073203e+03]
 [  0.00000000e+00   1.89405970e+01   1.67296973e+03]
 [  0.00000000e+00   0.00000000e+00   1.00000000e+00]]

[[  1.67129267e+00  -1.14127805e-01  -2.21808387e-01   2.40263807e-02
   -1.61877775e-03]]


10 + flags = 256 'fix intrinsics'
54.50464398008055

[[  1.50512671e+02   0.00000000e+00   1.08574478e+03]
 [  0.00000000e+00   9.14708563e+01   1.66977834e+03]
 [  0.00000000e+00   0.00000000e+00   1.00000000e+00]]

[[ 0.59014476 -0.0191317   0.09039141 -0.31884398 -0.00223853]]

'''