-
Notifications
You must be signed in to change notification settings - Fork 0
/
karol.py
265 lines (228 loc) · 9.93 KB
/
karol.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
import cv2
import numpy as np
import sys
from scipy.ndimage import rotate
import operator
def find_tree_contours(gray, param):
_,thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY_INV) # threshold
kernel = cv2.getStructuringElement(cv2.MORPH_CROSS,(3,3))
try:
dilated = cv2.dilate(thresh,kernel,iterations = param) # dilate
except:
dilated = cv2.dilate(thresh,kernel,iterations=1) # dilate
im2, contours, hierarchy = cv2.findContours(dilated, cv2.RETR_LIST, cv2.CHAIN_APPROX_TC89_KCOS)
outs = []
for i, contour in enumerate(contours):
[x,y,w,h] = cv2.boundingRect(contour)
if h < gray.shape[0] * 0.5 or w < gray.shape[1] * 0.5: continue
#cv2.rectangle(img, (x,y), (x+w,y+h), (255,0,255), 1)
out = np.zeros(gray.shape, dtype=np.uint8) + 255
cv2.drawContours(out, contours, i, 0, -1)
outs.append(out)
#out2[out < 2010] = 0
return outs
def find_tree_contour(gray, param, min_frac=0.5):
_,thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY_INV) # threshold
kernel = cv2.getStructuringElement(cv2.MORPH_CROSS,(3,3))
dilated = cv2.dilate(thresh,kernel,iterations = param) # dilate
im2, contours, hierarchy = cv2.findContours(dilated, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
out = np.zeros(gray.shape, dtype=np.uint8) + 255
for i, contour in enumerate(contours):
[x,y,w,h] = cv2.boundingRect(contour)
if gray.shape[-1] > gray.shape[-2]:
if w < gray.shape[-1] * min_frac:
continue
elif gray.shape[-1] < gray.shape[-2]:
if h < gray.shape[-2] * min_frac:
continue
else:
if h < gray.shape[-2] * min_frac or w < gray.shape[-1] * min_frac:
continue
#cv2.rectangle(img, (x,y), (x+w,y+h), (255,0,255), 1)
cv2.drawContours(out, contours, i, 0, -1)
out2 = np.zeros(gray.shape, dtype=np.uint8) + 255
out2[out == 0] = gray[out == 0]
return out2
def get_contour(img, gray, param=1):
tree_contour = find_tree_contour(gray, param)
dst = cv2.cornerHarris(tree_contour, 3, 3, 0.01)
dst = cv2.dilate(dst, None)
corners = dst > 0.05 * dst.max()
points = []
for y in range(img.shape[0]):
for x in range(img.shape[1]):
if corners[y,x]:
points.append((x,y))
return points
bottom_right = lambda N, M, k: np.fliplr(np.tri(N, M, k=k)) == 1
top_left = lambda N, M, k: np.flipud(np.tri(N, M, k=k)) == 1
bottom_left = lambda N, M, k: np.tri(N, M, k=k) == 1
top_right = lambda N, M, k: np.flipud(np.fliplr(np.tri(N, M, k=k))) == 1
def find_connected_nodes(image, i, nodes):
out = []
x, y = nodes[i]
for j in range(i + 1, len(nodes)):
if i == j: continue
distance = np.sqrt(np.sum(np.power(np.array(nodes[i]) - np.array(nodes[j]), 2)))
dst_x, dst_y = nodes[j]
angle = np.arctan2(dst_y - y, dst_x - x)
min_x = min(x, dst_x)
min_y = min(y, dst_y)
max_x = max(x, dst_x)
max_y = max(y, dst_y)
border = 12
A = image[min_y - border : max_y + border + 1, min_x - border : max_x + border + 1]
if np.min(A.shape) == 0:
continue
cutout = np.copy(A)
if ((x - min_x) == 0 and (y - min_y) == 0) or ((dst_x - min_x) == 0 and (dst_y - min_y) == 0):
cutout[top_right(cutout.shape[0], cutout.shape[1], min(cutout.shape) // 2 - cutout.shape[0])] = 255
cutout[bottom_left(cutout.shape[0], cutout.shape[1], min(cutout.shape) // 2 - cutout.shape[0])] = 255
else:
cutout[top_left(cutout.shape[0], cutout.shape[1], min(cutout.shape) // 2 - cutout.shape[0])] = 255
cutout[bottom_right(cutout.shape[0], cutout.shape[1], min(cutout.shape) // 2 - cutout.shape[0])] = 255
contours = find_tree_contours(cutout, 2)
for contour in contours:
other = False
for k in range(len(nodes)):
if k == j or k == i: continue
k_x, k_y = nodes[k]
k_x = k_x - min_x + border
k_y = k_y - min_y + border
if k_x >= border and k_x < cutout.shape[-1] - border and k_y >= border and k_y < cutout.shape[-2] - border:
if contour[k_y, k_x] < 200:
other = True
break
if other == False and contour[y - min_y + border, x - min_x + border] < 200 and contour[dst_y - min_y + border, dst_x - min_x + border] < 200:
out.append(j)
return out
def depth_first_cycle_finder(img, nodes, node_neighbors, original_root, root=None, parent=None, sources = {}):
if root is None:
sources = {}
sources[original_root] = original_root
root = original_root
for neighbor in node_neighbors[root]:
if neighbor == parent:
continue
if neighbor in sources:
# print("OOPS!", root, neighbor)
#print(sources)
path = []
i = root
while sources[i] != neighbor:
path.append(i)
i = sources[i]
path.append(i)
path.append(neighbor)
#print(path)
max_path = []
max_d = 0
for j in range(len(path)):
d = np.sqrt(np.sum(np.power(np.array(nodes[path[j]]) - np.array(nodes[path[j-1]]), 2)))
if d > max_d:
max_path = (path[j], path[j - 1])
max_d = d
a, b = max_path
node_neighbors[a] = [x for x in node_neighbors[a] if x != b]
node_neighbors[b] = [x for x in node_neighbors[b] if x != a]
return None
sources[neighbor] = root
#print(root, neighbor)
if depth_first_cycle_finder(img, nodes, node_neighbors, original_root, neighbor, root, sources) == None:
return None
return sources
# Nodes from corners
def nodes_from_corners(image, gray, points, max_dist=16, iterations=1):
if iterations == 0:
out = []
for j, point in enumerate(points):
x, y = map(int, point)
A = gray[y - 1 : y + 2, x - 1 : x + 2]
off_y, off_x = np.array(np.unravel_index(A.argmin(), A.shape)) - 1
new_x, new_y = x + off_x, y + off_y
if gray[new_y, new_x] < gray[y, x]:
out.append((new_x,new_y))
else:
out.append((x,y))
return out
# print('iternations:', iterations)
nodes = []
while len(points):
point = points[-1]
points.pop()
neighbors = [point]
for i in reversed(range(len(points))):
dist = np.sqrt(np.sum(np.power(np.array(point) - np.array(points[i]), 2)))
if dist < max_dist:
neighbors.append(points[i])
points.pop(i)
neighbors = np.array(neighbors)
point = neighbors[np.random.randint(len(neighbors))]
point = np.mean(neighbors, axis=0)
# print('Remaining corner points to assign a corner:', len(points))
if iterations == 1:
#image[point[1], point[0], :] = (255, 0, 0)
cv2.circle(image, tuple(map(int, point)), 4, (255,128,0), 2)
nodes.append(point)
return nodes_from_corners(image, gray, nodes, max_dist + 2, iterations - 1)
def node_neighbors(nodes, gray):
# Figure out node neighbors
node_neighbors = {}
for i, node in enumerate(nodes):
node_neighbors[i] = find_connected_nodes(gray, i, nodes)
# Do reciprocal adding of nodes
for i in range(len(nodes)):
for j in node_neighbors[i]:
if i not in node_neighbors[j]:
node_neighbors[j].append(i)
return node_neighbors
def unhook_triangles(nodes, node_neighbors):
for i in range(len(nodes)):
for j in range(i + 1, len(nodes)):
for k in range(j + 1, len(nodes)):
if (i in node_neighbors[j] and i in node_neighbors[k] and
j in node_neighbors[i] and j in node_neighbors[k] and
k in node_neighbors[i] and k in node_neighbors[j]):
d1 = np.sqrt(np.sum(np.power(np.array(nodes[i]) - np.array(nodes[j]), 2)))
d2 = np.sqrt(np.sum(np.power(np.array(nodes[i]) - np.array(nodes[k]), 2)))
d3 = np.sqrt(np.sum(np.power(np.array(nodes[j]) - np.array(nodes[k]), 2)))
max_d = max(d1, d2, d3)
if d1 == max_d:
node_neighbors[i] = [x for x in node_neighbors[i] if x != j]
node_neighbors[j] = [x for x in node_neighbors[j] if x != i]
elif d2 == max_d:
node_neighbors[i] = [x for x in node_neighbors[i] if x != k]
node_neighbors[k] = [x for x in node_neighbors[k] if x != i]
else:
node_neighbors[j] = [x for x in node_neighbors[j] if x != k]
node_neighbors[k] = [x for x in node_neighbors[k] if x != j]
# print('Triangle', i, j, k)
def remove_spurious(nodes, node_neighbors):
for i in range(len(nodes)):
if len(node_neighbors[i]) == 2:
n1, n2 = node_neighbors[i]
for j in node_neighbors:
#print(node_neighbors[j], i, j)
node_neighbors[j] = [x for x in node_neighbors[j] if x != i]
node_neighbors[n1].append(n2)
node_neighbors[n2].append(n1)
del node_neighbors[i]
def find_root(nodes, node_neighbors):
leaves = []
for i in node_neighbors:
if len(node_neighbors[i]) == 1:
leaves.append(i)
max_D = 0
max_i = 0
for i in leaves:
min_d = 1000000
for j in leaves:
if i == j:
continue
d = np.sqrt(np.sum(np.power(np.array(nodes[i]) - np.array(nodes[j]), 2)))
if d < min_d:
min_d = d
if min_d > max_D:
max_i = i
max_D = min_d
return max_i