-
Notifications
You must be signed in to change notification settings - Fork 0
/
detect_rectangles.py
202 lines (159 loc) · 7.99 KB
/
detect_rectangles.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
import os
import numpy as np
import cv2 as cv
import uuid
from scipy.spatial import ConvexHull
import json_exporter
import copy
from typing import Tuple
cv2 = cv
# TODO don't hardcode
original_interval = {'x': (0, 16), 'y': (0, 9), 'z': (0, 4)}
def morph(input_img):
kernel = np.ones((21, 21), np.uint8)
morph_img = cv2.dilate(input_img, kernel, iterations=1)
morph_img = cv2.morphologyEx(morph_img, cv2.MORPH_OPEN, kernel)
morph_img = cv2.morphologyEx(morph_img, cv2.MORPH_CLOSE, kernel)
kernel = np.ones((20, 20), np.uint8)
morph_img = cv2.dilate(morph_img, kernel, iterations=1)
return morph_img
def separate_wall(img_:np.ndarray, threshold:np.float=3.0)-> Tuple[np.ndarray, np.ndarray]:
"""
Naive wall vs obstacles separations
Parameters
----------
img_ : the image input matrix
threshold : percentage threshold to define wall vs everytihng else
Returns : Tuple[obstacles:np.ndarray, walls: np.ndarray]
-------
TODO fix the distance of wall to the bounding box!
"""
img_original = copy.deepcopy(img_)
img_wall = copy.deepcopy(img_) * 0
img_obstacles = copy.deepcopy(img_)
def point_to_line(point, p1, p2):
# line between p1, p2
# check calculate shapes
assert p1.shape[0] == 2
assert p2.shape[0] == 2
assert point.shape[1] == 2
zaeler = np.abs((p2[1] - p1[1]) * point[:, 0] - (p2[0] - p1[0]) * point[:, 1] + p2[0] * p1[1] - p2[1] * p1[0])
nenner = np.sqrt(np.square(p2[1] - p1[1]) + np.square(p2[0] - p1[0]))
return zaeler / nenner
point_coords_original_img = np.where(img_original > 0)
point_coords_original_img = np.hstack([point_coords_original_img[0].reshape([-1, 1]),
point_coords_original_img[1].reshape([-1, 1])])
hull = ConvexHull(point_coords_original_img)
hull_points = point_coords_original_img[hull.vertices]
for ix in range(hull_points.shape[0]):
start = hull_points[ix, :]
if (ix + 1) == hull_points.shape[0]:
end = hull_points[0, :]
else:
end = hull_points[ix + 1, :]
distances = point_to_line(point_coords_original_img, start, end)
#thresh = np.percentile(distances, threshold) # 0.001 - one promile
zero_point_coords = point_coords_original_img[np.where(distances < 80)]
img_obstacles[zero_point_coords[:, 0], zero_point_coords[:, 1]] = 0 # inverted?
img_wall[zero_point_coords[:, 0], zero_point_coords[:, 1]] = img_original[zero_point_coords[:, 0], zero_point_coords[:, 1]]
return img_obstacles, img_wall
# height map
background_img_path = 'data/entire_hall.png'
img_original_ = cv.imread(background_img_path, 0)
img_obstacles, img_wall = separate_wall(img_original_, 1.5)
# TODO set this to run with the correct input depending on what you would like to use
img_original = img_obstacles
img_original = img_wall
img_all_colors = cv2.cvtColor(img_original, cv2.COLOR_GRAY2RGB)
img_original = morph(img_original)
# TODO find color thresholds using kmeans
all_contours = []
color_interval_len = 32
remove_last_interval = 1
for color_interval_start in range(color_interval_len, 255 - remove_last_interval * color_interval_len,
color_interval_len):
img = cv.inRange(img_original, color_interval_start, color_interval_start + color_interval_len)
contours, hierarchy = cv.findContours(img, cv2.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
# cv.drawContours(img, contours, -1, (0,255,0), 3)
#
# cv.imwrite('data/result.png', img)
img_cnt = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
img_outer_cnt = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
img_approx = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
img_hulls = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
img_area_filtered = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
img_cnt_filled = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
#
def create_graph(vertex, color, image):
for g in range(0, len(vertex) - 1):
for y in range(0, len(vertex[0][0]) - 1):
cv.circle(image, (vertex[g][0][y], vertex[g][0][y + 1]), 3, (255, 255, 255), -1)
cv.line(image, (vertex[g][0][y], vertex[g][0][y + 1]), (vertex[g + 1][0][y], vertex[g + 1][0][y + 1]),
color, 2)
cv.line(image, (vertex[len(vertex) - 1][0][0], vertex[len(vertex) - 1][0][1]),
(vertex[0][0][0], vertex[0][0][1]),
color, 2)
#
# for b,cnt in enumerate(contours):
# if hierarchy[0,b,3] == -1: #<-the mistake might be here
# approx = cv.approxPolyDP(cnt,0.015*cv.arcLength(cnt,True), True)
# clr = (255, 0, 0)
# create_graph(approx, clr) #function for drawing the found contours in the new img
# cv.imwrite('starg.jpg', newimg)
hierarchy = hierarchy[0] # get the actual inner list of hierarchy descriptions
if not isinstance(hierarchy, np.ndarray) or len(hierarchy.shape) == 1:
continue
# For each contour, find the bounding rectangle and draw it
for currentContour, currentHierarchy in zip(contours, hierarchy):
# x, y, w, h = cv2.boundingRect(currentContour)
# if currentHierarchy[2] < 0:
# # these are the innermost child components
# # cv2.rectangle(newimg, (x, y), (x + w, y + h), (0, 0, 255), 20)
# pass
# elif currentHierarchy[3] < 0:
# # these are the outermost parent components
# cv2.rectangle(newimg, (x, y), (x + w, y + h), (0, 255, 0), 20)
create_graph(currentContour, (0, 0, 255), img_cnt)
cv.fillPoly(img_cnt_filled, pts=[currentContour], color=(0, 255, 0))
if currentHierarchy[3] < 0:
create_graph(currentContour, (0, 0, 255), img_outer_cnt)
approx = cv.approxPolyDP(currentContour, 0.001 * cv.arcLength(currentContour, True), True)
create_graph(approx, (0, 0, 255), img_approx)
hull = approx
create_graph(hull, (0, 0, 255), img_hulls)
area = cv.contourArea(approx)
if area >= 10000:
create_graph(hull, (0, 0, 255), img_area_filtered)
create_graph(hull, (0, 0, 255), img_all_colors)
col_range_end = color_interval_start + color_interval_len
all_contours.append((currentContour, col_range_end))
# Finally show the image
if not os.path.exists('data/steps'):
os.makedirs('data/steps')
cv.imwrite(f'data/steps/img_r{color_interval_start}_1_cnt.png', img_cnt)
cv.imwrite(f'data/steps/img_r{color_interval_start}_2_outer_cnt.png', img_outer_cnt)
cv.imwrite(f'data/steps/img_r{color_interval_start}_3_approx.png', img_approx)
cv.imwrite(f'data/steps/img_r{color_interval_start}_4_hulls.png', img_hulls)
cv.imwrite(f'data/steps/img_r{color_interval_start}_5_area_filtered.png', img_area_filtered)
cv.imwrite(f'data/steps/img_r{color_interval_start}_5_img_all_colors.png', img_all_colors)
cv.imwrite(f'data/steps/img_r{color_interval_start}_5_img_cnt_filled.png', img_cnt_filled)
cv.imwrite(f'data/steps/img_final_all_contours.png', img_all_colors)
def map_to_interval(val, old_interval, new_interval):
A, B = old_interval
a, b = new_interval
return (val - A) * (b - a) / (B - A) + a
interval_x = (0, img_original.shape[1])
interval_y = (0, img_original.shape[0])
layers_json_format = []
for contour, height_intensity in all_contours:
contour_points = []
for i, point in enumerate(contour):
contour_points.append({'x': round(map_to_interval(point[0][0], interval_x, original_interval['x']), 2),
'y': round(map_to_interval(point[0][1], interval_y, original_interval['y']), 2), 'id': i + 1})
layers_json_format.append(
{'points': contour_points,
'height': round(map_to_interval(height_intensity, (0, 255), original_interval['z']), 2),
'shape_type': 'obstacle',
'shapeId': str(uuid.uuid1())
})
json_exporter.export('data/result.json', layers_json_format, background_img_path)