Newer
Older
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
import datetime
import os
import json
import sys
import argparse
import cv2
import numpy as np
import utils
import solve_lens
def parse_arguments():
usage_text = (
"Usage: python aruco-frame.py [options]"
)
parser = argparse.ArgumentParser(description=usage_text)
parser.add_argument("-i", "--input", type=str,
help="Input filename.")
parser.add_argument("-o", "--output", type=str, default="",
help="Output filename (default: <filename_in>_extracted.png).")
parser.add_argument("-d", "--dpi", type=int, default=-1,
help="Manual output DPI (default: auto).")
parser.add_argument("-s", "--show", action="store_true",
help="Show debug image.")
parser.add_argument("-c", "--config", type=str, default="./config/config.json",
help="Frame configuration file (default: ./config/config.json).")
parser.add_argument("-v", "--verbose", action="store_true",
help="Verbose mode (default: false).")
return parser.parse_args()
def imshow(img, h_view=700, win_name="debug"):
h, w = img.shape[:2]
w_view = int(h_view * w / h)
cv2.imshow(win_name, cv2.resize(img, (w_view, h_view), interpolation=cv2.INTER_AREA))
cv2.waitKey(0)
def extract_image(img, proj, config, dots_per_mm, dist_params=None):
h, w, c = img.shape
m = config["margins"]["inner_content"]
xmin = m
xmax = config["width"] - m
ymin = m
ymax = config["height"] - m
h_out = int(dots_per_mm * (ymax - ymin))
w_out = int(dots_per_mm * (xmax - xmin))
x = np.linspace(xmin, xmax, w_out)
y = np.linspace(ymax, ymin, h_out)
xx, yy = np.meshgrid(x, y)
xy_list = np.ones((h_out * w_out, 2))
xy_list[:, 0] = xx.flatten()
xy_list[:, 1] = yy.flatten()
uv_src = apply_affine(proj, xy_list)
if dist_params is not None:
k1, k2, uc, vc = dist_params[:]
mat = np.array([[w, 0, uc], [0, w, vc], [0, 0, 1]], dtype=np.float32)
dist_coeffs = np.array([[0, 0, 0, 0, 0, k1, k2, 0]], dtype=np.float32)
out = cv2.undistortPoints(uv_src, mat, dist_coeffs, P=mat)
uv_src = out[:, 0, :]
map1 = uv_src[:, 0].reshape((h_out, w_out)).astype(np.float32)
map2 = uv_src[:, 1].reshape((h_out, w_out)).astype(np.float32)
# plt.figure()
# plt.imshow(map1)
# plt.figure()
# plt.imshow(map2)
# plt.show()
img_out = cv2.remap(img, map1, map2, interpolation=cv2.INTER_CUBIC)
return img_out
def find_aruco(img):
aruco_dict = cv2.aruco.getPredefinedDictionary(cv2.aruco.DICT_4X4_50)
params = cv2.aruco.DetectorParameters()
params.adaptiveThreshWinSizeMax = 40
params.useAruco3Detection = True
corners, ids, rejected = cv2.aruco.detectMarkers(img, dictionary=aruco_dict, parameters=params)
if ids is None:
return {}
else:
corners_dict = {ids[k][0]: corners[k][0, :, :] for k in range(len(ids))}
return corners_dict
def identify_frame(img, config_frames, debug=False):
corners_dict = find_aruco(img)
if debug:
img_view = np.copy(img)
for c in corners_dict.values():
for uv in c[:, :]:
cv2.circle(img_view, uv.astype(np.int32), radius=30, color=(0, 0, 255), thickness=cv2.FILLED)
imshow(img_view)
name_found = None
for name in config_frames:
match = True
for aruco_id in config_frames[name]["aruco_id"]:
if aruco_id not in corners_dict:
match = False
break
if match:
name_found = name
break
return name_found
def get_aruco_features(img, config):
corners_dict_all = find_aruco(img)
corners_dict = {k: corners_dict_all[k] for k in config["aruco_id"]}
centers_dict = {k: np.mean(corners_dict[k], axis=0) for k in corners_dict}
xy_array = np.zeros((4, 2))
uv_array = np.zeros((4, 2))
for i in range(4):
xy_array[i, :] = config["aruco_pos"][i]
uv_array[i, :] = centers_dict[config["aruco_id"][i]]
return xy_array, uv_array
def apply_affine(a, xy):
n = len(xy)
xyz = np.ones((n, 3))
xyz[:, :2] = xy
uvw = xyz @ a.T
uv = uvw[:, :2] / uvw[:, 2:]
return uv
def get_corner_features(img_gray, proj, config):
n_points = sum(len(edge) for edge in config["corner_pos"])
xy_feats = np.zeros((n_points, 2))
uv_feats_approx = np.zeros((n_points, 2))
k = 0
for edge in config["corner_pos"]:
n_edge = len(edge)
xy_feats[k:k + n_edge, :] = np.array(edge)
uv_feats_approx[k:k + n_edge] = apply_affine(proj, xy_feats[k:k + n_edge, :])
k += n_edge
# adjust search region to resolution
search_mm = 0.7 * config["corner_size"] / 2
cross_xy = np.zeros((4 * n_points, 2), dtype=np.float32)
cross_xy[0::4, :] = xy_feats - np.array([search_mm, 0])
cross_xy[1::4, :] = xy_feats + np.array([search_mm, 0])
cross_xy[2::4, :] = xy_feats - np.array([0, search_mm])
cross_xy[3::4, :] = xy_feats + np.array([0, search_mm])
cross_uv = apply_affine(proj, cross_xy)
cross_uv_r = cross_uv.reshape(n_points, 4, 2)
span_uv = (np.max(cross_uv_r, axis=1) - np.min(cross_uv_r, axis=1)) / 2
search_uv = np.mean(span_uv, axis=0).astype(np.int32)
# print(search_uv)
criteria = (cv2.TERM_CRITERIA_COUNT + cv2.TERM_CRITERIA_EPS, 40, 0.001)
ret = cv2.cornerSubPix(img_gray,
uv_feats_approx[:, np.newaxis, :].astype(np.float32),
(search_uv[0], search_uv[1]),
(-1, -1),
criteria)
uv_feats = ret[:, 0, :]
# return xy_feats, uv_feats_approx
return xy_feats, uv_feats
def get_dots_per_mm(xy, uv, use_max=True):
xy_dist = np.zeros((4,))
uv_dist = np.zeros((4,))
for i in range(-1, 3):
xy_dist[i] = np.linalg.norm(xy[i + 1] - xy[i])
uv_dist[i] = np.linalg.norm(uv[i + 1] - uv[i])
if use_max:
return np.max(uv_dist / xy_dist)
else:
return np.mean(uv_dist / xy_dist)
def process_image(img, config_frames, solve_dist=False, view=False, view_radius=16, verbose=False, dpi=None):
h, w = img.shape[:2]
if len(img.shape) == 2:
img_gray = img
img_rgb = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
else:
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img_rgb = img
frame_name = identify_frame(img_rgb, config_frames)
if verbose:
print(f"frame found: '{frame_name}'")
if frame_name is None:
raise RuntimeError("No frame found!")
config = config_frames[frame_name]
xy_a, uv_a = get_aruco_features(img_rgb, config)
proj = utils.solve_affine(xy_a, uv_a)
if dpi is None:
dpi = int(get_dots_per_mm(xy_a, uv_a) * 25.4)
dots_per_mm = dpi / 25.4
if verbose:
print(f"DPI: {dpi}")
xy_c, uv_c = get_corner_features(img_gray, proj, config)
proj_fine = utils.solve_affine(xy_c, uv_c)
err1 = solve_lens.xy_error(xy_c, uv_c, proj)
err2 = solve_lens.xy_error(xy_c, uv_c, proj_fine)
if verbose:
print(f"Error init. : {np.mean(np.linalg.norm(err1, axis=1)):.3f} mm")
print(f"Error refined : {np.mean(np.linalg.norm(err2, axis=1)):.3f} mm")
if view:
img_view = np.copy(img_rgb)
for uv in uv_a:
cv2.circle(img_view, uv.astype(np.int32), radius=view_radius, color=(255, 200, 0), thickness=cv2.FILLED)
for uv in uv_c:
cv2.circle(img_view, uv.astype(np.int32), radius=view_radius, color=(0, 0, 255), thickness=cv2.FILLED)
imshow(img_view, win_name="features")
# cv2.imwrite("view.jpg", img_view)
if solve_dist:
params = solve_lens.solve_distortion(xy_c, uv_c, proj_fine, w, w, h)
for i in range(4):
uv_u = solve_lens.undistort(params, uv_c, w)
proj_fine = utils.solve_affine(xy_c, uv_u)
params = solve_lens.solve_distortion(xy_c, uv_c, proj_fine, w, w, h)
uv_u = solve_lens.undistort(params, uv_c, w)
err3 = solve_lens.xy_error(xy_c, uv_u, proj_fine)
if verbose:
print(f"Error lens dist.: {np.mean(np.linalg.norm(err3, axis=1)):.3f} mm")
img_out = extract_image(img_rgb, proj_fine, config, dots_per_mm, dist_params=params)
else:
img_out = extract_image(img_rgb, proj_fine, config, dots_per_mm)
if view:
imshow(img_out, win_name="out")
# handle upside down case
if uv_a[0][1] < uv_a[2][1]:
img_out = cv2.rotate(img_out, cv2.ROTATE_180)
h_out, w_out, _ = img_out.shape
if verbose:
print(f"Dots per mm: {dots_per_mm:.2f}")
print(f"Dots per in: {dpi}")
print(f"Resolution: {w_out} x {h_out}")
return img_out, dpi
def load_config_frames(filename):
head, tail = os.path.split(filename)
with open(filename, "r") as f:
config_all = json.load(f)
config = {}
for frame_name, frame_filename in config_all.items():
with open(os.path.join(head, frame_filename), "r") as f:
config[frame_name] = json.load(f)
return config
def main():
args = parse_arguments()
filename_in = args.input
if args.output == "":
filename_out = os.path.splitext(filename_in)[0] + "_extracted.png"
else:
filename_out = args.output
head_out, _ = os.path.split(filename_out)
print(f"Processing: '{filename_in}' -> '{filename_out}'")
img = cv2.imread(filename_in, cv2.IMREAD_UNCHANGED)
config_frames = load_config_frames(args.config)
if args.dpi == -1:
img_out, dpi = process_image(img, config_frames,
solve_dist=True, view=args.show, verbose=args.verbose)
else:
img_out, dpi = process_image(img, config_frames,
solve_dist=True, view=args.show, verbose=args.verbose, dpi=args.dpi)
utils.writePNGwithdpi(filename_out, img_out, dpi=(dpi, dpi))
print("Done.")
if __name__ == "__main__":
main()