-
Notifications
You must be signed in to change notification settings - Fork 9
/
template_matching.py
128 lines (112 loc) · 5.09 KB
/
template_matching.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
import cv2
import numpy as np
from matplotlib import pyplot as plt
import copy
import time
def rotate_and_scale(image, scale=1.0, angle=0):
height, width = image.shape[:2]
mat = cv2.getRotationMatrix2D(center=(width / 2, height / 2), angle=angle, scale=scale)
new_w, new_h = width * scale, height * scale
r = np.deg2rad(angle)
new_w, new_h = (abs(np.sin(r) * new_h) + abs(np.cos(r) * new_w), abs(np.sin(r) * new_w) + abs(np.cos(r) * new_h))
(tx, ty) = ((new_w - width)/2., (new_h - height)/2.)
mat[0, 2] += tx
mat[1, 2] += ty
rotated_image = cv2.warpAffine(src=image, M=mat, dsize=(int(new_w), int(new_h)))
return rotated_image, mat
img = cv2.imread('./test-auto-cropped/2.bmp', 0)
# img = cv2.resize(src=img, dsize=(1682, 606))
rotated_img = copy.deepcopy(x=img)
print('image shape:', img.shape)
template = cv2.imread('./test-images/t7_cropped.jpg', 0)
print('template shape:', template.shape)
template_h, template_w = template.shape[:2]
img_h, img_w = img.shape[:2]
mid_h, mid_w = int(img_h/2.), int(img_w/2.)
"""best value template matching"""
# res = cv2.matchTemplate(img, template, cv2.TM_CCOEFF_NORMED)
# min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(res)
# top_left = max_loc
# bottom_right = (top_left[0] + template_w, top_left[1] + template_h)
# cv2.rectangle(img, top_left, bottom_right, 0, 2)
"""end of best value template matching"""
"""multiple values"""
# threshold = 0.55
# loc = np.where(res >= threshold)
# for pt in zip(*loc[::-1]):
# cv2.rectangle(img_gray, pt, (pt[0] + w, pt[1] + h), (0, 0, 255), 2)
"""end of multiple values"""
"""scale independent"""
# found = None
# for scale in np.linspace(1, 2, 4)[::-1]:
# resized_template = cv2.resize(template, (int(template_w * scale), int(template_h * scale)))
# # ratio = float(resized_template.shape[1]) / template_w
# if resized_template.shape[0] > img_h or resized_template.shape[1] > img_w:
# continue
# res = cv2.matchTemplate(img, resized_template, cv2.TM_CCOEFF_NORMED)
# min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(res)
# if found is None or max_val > found[0]:
# found = (max_val, max_loc)
# print('found is:', found)
# (_, max_loc) = found
# start_point = (int(max_loc[0] * scale), int(max_loc[1] * scale))
# end_point = (int((max_loc[0] + template_w) * scale), int((max_loc[1] + template_h) * scale))
# cv2.rectangle(img, start_point, end_point, 0, 2)
"""end of scale independent"""
"""rotation and scale invariant"""
t0 = time.time()
found = None
for degree in np.linspace(start=0, stop=360, num=46):
# rotated_template, M = rotate_and_scale(image=template, scale=1.0, angle=degree)
rotated_img, M = rotate_and_scale(image=rotated_img, scale=1.0, angle=degree)
# if rotated_template.shape[0] > img_h or rotated_template.shape[1] > img_w:
# continue
"""single"""
res = cv2.matchTemplate(rotated_img, template, cv2.TM_CCOEFF_NORMED)
min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(res)
found = (max_val, max_loc)
print('found is:', found)
print('degree is:', degree)
if max_val >= 0.7:
org_img = copy.deepcopy(x=img)
rotated_org_img, _ = rotate_and_scale(image=org_img, scale=1.0, angle=degree)
sp = (int(max_loc[0]), int(max_loc[1]))
ep = (int(max_loc[0] + template_w), int(max_loc[1] + template_h))
cv2.rectangle(rotated_img, sp, ep, 0, 2)
plt.imshow(X=rotated_org_img[sp[1]:ep[1], sp[0]:ep[0]], cmap='gray')
plt.show()
big_img, _ = rotate_and_scale(image=rotated_img, scale=1.0, angle=(360 - degree))
big_img_h, big_img_w = big_img.shape[:2]
rotated_img = big_img[int((big_img_h / 2.) - mid_h)+1:int((big_img_h / 2.) + mid_h),
int((big_img_w / 2.) - mid_w)+1:int((big_img_w / 2.) + mid_w)]
# plt.imshow(X=rotated_img, cmap='gray')
# plt.show()
#
# """multiple"""
# res = cv2.matchTemplate(img, rotated_template, cv2.TM_CCOEFF_NORMED)
# threshold = 0.7
# loc = np.where(res >= threshold)
# for pt in zip(*loc[::-1]):
# cv2.rectangle(img, pt, (pt[0] + template_w, pt[1] + template_h), (0, 0, 255), 2)
#
# top_left = sp
# top_right = (sp[0] + template_w, sp[1] + template_h)
# bottom_right = ep
# center = (np.mean(a=[top_left[0], bottom_right[0]]), np.mean(a=[top_left[1], bottom_right[1]]))
# top_left = (top_left[0] - center[0], top_left[1] - center[1])
# top_right = (top_right[0] - center[0], top_right[1] - center[1])
# bottom_right = (bottom_right[0] - center[0], bottom_right[1] - center[1])
# rotation_matrix = M[:, 0:2]
# top_left = np.array([[top_left[0]], [top_left[1]]])
# top_right = np.array([[top_right[0]], [top_right[1]]])
# bottom_right = np.array([[bottom_right[0]], [bottom_right[1]]])
#
# top_left = np.matmul(rotation_matrix, top_left)
# top_right = np.matmul(rotation_matrix, top_right)
# bottom_right = np.matmul(rotation_matrix, bottom_right)
"""end of rotation and scale invariant"""
t1 = time.time()
print('elapsed time: %.2f' % (t1-t0))
plt.imshow(rotated_img, cmap='gray')
plt.show()
# cv2.imwrite('./results/ritm_final.bmp', rotated_img)