-
Notifications
You must be signed in to change notification settings - Fork 8
/
Copy pathsurface.py
332 lines (263 loc) · 12.6 KB
/
surface.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
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
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
""" This module contains functions and structures for detecting and transforming the playing surface """
# Import standard libraries
import numpy as np
import cv2
import imutils
from copy import deepcopy
from datetime import datetime
# Set the cutoff limit for detected surfaces
# e.g. 0.5 means a surface is only valid if it occupies at
# least half of the entire image space from the original image
cutoff = 0.03
class PlayingSurface:
" Structure to store information about the playing surface. "
def __init__(self):
self.name = "playing_surface_name"
self.transform = [] # Transformed image of the playing surface
self.contour = [] # Contour of the playing surface wrt original image
self.img_cnt = [] # Image with contour overlayed on original image
self.area = [] # Area of the playing surface
self.area_relative = [] # Relative size of the playing surface wrt original image
self.perspective_matrix = []
self.width = []
self.height = []
self.dealer_region = []
self.player_region = []
def detect(image):
" This function finds the playing surface in the original image and stores info in an appropriate object "
# The image will be set to this height for faster processing
image_resize_value = 300.0
# Work out how much bigger than 'x' the image is wrt height
ratio = image.shape[0] / image_resize_value
# Make a copy of the original image at its full size
original_image = deepcopy(image)
# Resize the image (should be a reduction, but not strictly important)
image = imutils.resize(image, height=int(image_resize_value))
# Convert the image to grayscale, apply a blur and then find edges
im_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
im_blur = cv2.bilateralFilter(im_gray, 11, 17, 17)
im_edge = cv2.Canny(im_blur, 30, 200)
# Find contours in the edged image and only keep the largest five (ordered largest to smallest)
_, contours, _ = cv2.findContours(im_edge.copy(), cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
contours = sorted(contours, key=cv2.contourArea, reverse=True)[:5]
# Initialise an index to represent the location of the playing surface in the contours list
contour_idx = 0
# Initialise the polymetric approximation accuracy scaling factor
POLY_ACC_CONST = 0.02
surface_cnt = None
for c in contours:
# Approximate the contour
perimeter = cv2.arcLength(c, closed=True)
curr_cnt = cv2.approxPolyDP(c, POLY_ACC_CONST * perimeter, closed=True)
# If our approximated contour has four points, then we can assume that we have found our target
# because the contours are ordered largest to smallest and we want the largest four-point contour
if len(curr_cnt) == 4:
# Draw the contour over the image (to be displayed later)
# image = im_edge
cv2.drawContours(image, contours, contourIdx=contour_idx, color=(255, 180, 0), thickness=2)
surface_cnt = curr_cnt
break
# Increment the index
contour_idx += 1
# Even if there're contours but none of them are 4 points, we want to
# return
if surface_cnt is None:
return None
# Now that we have our contour, we need to determine the top-left, top-right, bottom-right, and
# bottom-left points so that we can later warp the image -- we'll start by reshaping our contour
# to be our finals and initializing our output rectangle in top-left, top-right, bottom-right,
# and bottom-left order
rect_points = surface_cnt.reshape(4, 2)
# The top-left point has the smallest sum whereas the bottom-right has the largest sum
s = np.sum(rect_points, axis=1)
top_left_point = rect_points[np.argmin(s)]
bottom_right_point = rect_points[np.argmax(s)]
temp = None
reduced_points = None
for i in range(rect_points.shape[0]):
if np.array_equal(rect_points[i], top_left_point):
temp = np.delete(rect_points, i, 0)
break
for i in range(temp.shape[0]):
if np.array_equal(temp[i], bottom_right_point):
reduced_points = np.delete(temp, i, 0)
break
# Compute the difference between the points -- the top-right will have the minimum difference
# and the bottom-left will have the maximum difference
diff = np.diff(reduced_points, axis=1)
top_right_point = reduced_points[np.argmin(diff)]
bottom_left_point = reduced_points[np.argmax(diff)]
# Scale points up to full size
top_left_point = np.multiply(top_left_point, ratio)
top_right_point = np.multiply(top_right_point, ratio)
bottom_right_point = np.multiply(bottom_right_point, ratio)
bottom_left_point = np.multiply(bottom_left_point, ratio)
# Put all the points into a single 4 x 2 ndarray
rect = np.array([top_left_point, top_right_point, bottom_right_point, bottom_left_point], dtype="float32")
# Compute the width of the new image
widthA = np.sqrt(((bottom_right_point[0] - bottom_left_point[0]) ** 2) +
((bottom_right_point[1] - bottom_left_point[1]) ** 2))
widthB = np.sqrt(((top_right_point[0] - top_left_point[0]) ** 2) +
((top_right_point[1] - top_left_point[1]) ** 2))
width = max(int(widthA), int(widthB))
# Compute the height of the new image
heightA = np.sqrt(((top_right_point[0] - bottom_right_point[0]) ** 2) +
((top_right_point[1] - bottom_right_point[1]) ** 2))
heightB = np.sqrt(((top_left_point[0] - bottom_left_point[0]) ** 2) +
((top_left_point[1] - bottom_left_point[1]) ** 2))
height = max(int(heightA), int(heightB))
height = 800
width = int(1.8 * height)
# Configure the four points of the destination image for the transform function
# Noting that the origin is the top-left point with x positive to the right and
# y positive downwards
dst = np.array([
# top-left point
[0, 0],
# top-right point
[width - 1, 0],
# bottom-right point
[width - 1, height - 1],
# bottom-left point
[0, height - 1]], dtype="float32")
# calculate the perspective transform matrix and warp the perspective to grab
# a birds-eye view of the surface
# Noting that rect and dst must have their points in the same order
persp_mtx = cv2.getPerspectiveTransform(rect, dst)
transformed = cv2.warpPerspective(original_image, persp_mtx, (width, height))
# Get the area of the surface (pixels^2)
transformed_area = cv2.contourArea(surface_cnt)
# Get the area of the original image for a comparison (pixels^2)
original_image_area = original_image.shape[0] * original_image.shape[1]
# Compute the relative size of the surface wrt the entire image
relative_size = np.divide(transformed_area, original_image_area)
if relative_size < cutoff:
return None
else:
# Create a new instance of the playing surface class
playing_surface = PlayingSurface()
# Give it a name
playing_surface.name = 'primary'
# Store the contour of the playing surface
playing_surface.contour = surface_cnt
# Store the contour drawn over the image
playing_surface.img_cnt = image
# Store the transformed playing surface
playing_surface.transform = transformed
# Store the area of the surface (pixels^2)
playing_surface.area = transformed_area
# Store the relative size of the playing surface wrt the entire image
playing_surface.area_relative = relative_size
# Store the perspective matrix used to get the transformation
playing_surface.perspective_matrix = persp_mtx
# Store the width of the transformed image
playing_surface.width = width
# Store the height of the transformed image
playing_surface.height = height
playing_surface.dealer_region = np.array([0, int(width / 2)])
playing_surface.player_region = np.array([int((width / 2) + 1), int(width)])
# Return the complete playing surface object
return playing_surface
def display(countdown, contoured, transformed=np.array([])):
# Arbitrary x, y offsets for displays
x_offset = 50
y_offset = 50
# Original image from source with countdown timer
cv2.imshow("Original", countdown)
cv2.moveWindow("Original", x_offset, y_offset)
# Original image with the contour of the playing surface overlayed
cv2.imshow("Contoured", contoured)
cv2.moveWindow("Contoured", countdown.shape[1] + x_offset, y_offset)
# Only show the transformed surface if it exists (given to function implies existence)
if bool(transformed.any()):
# Transformed playing surface
cv2.imshow("Transformed", transformed)
cv2.moveWindow("Transformed", countdown.shape[1] * 2 + x_offset, y_offset)
return
def timer(image, count):
text = format(count, '02d')
font_pos = (380, 70)
# Give the font a little shadow to help it stand out
shadow_offset = 2
a = int(font_pos[0]) + shadow_offset
b = int(font_pos[1]) + shadow_offset
font_pos2 = (a, b)
font = cv2.FONT_HERSHEY_SIMPLEX
font_scale = 2
font_thickness = 3
line_type = cv2.LINE_AA
# Make the font red near the end of the countdown (white otherwise)
if count <= 10 and count % 2 == 0:
colour = (0, 0, 255)
elif count <= 3:
colour = (0, 0, 255)
else:
colour = (255, 255, 255)
# Add text to the image with a small shadow
cv2.putText(image, text, font_pos2, font, font_scale, (0, 0, 0), font_thickness, line_type)
cv2.putText(image, text, font_pos, font, font_scale, colour, font_thickness, line_type)
# Return the image with the text overlayed
return image
def get_surface(cap, count):
count += 1
curr_time = float(datetime.now().strftime('%s.%f')[9:-5])
# initialise an empty playing surface
valid_surface = None
while count != 0:
# Initialise or reinitialise the surface each time we grab a new frame
playing_surface = None
# take camera image
(flag_, image) = cap.read()
# Try and obtain a valid playing surface object
playing_surface = detect(image)
# Configure the raw original image
orig_disp = deepcopy(imutils.resize(image, height=300))
# Start with a fresh copy of the raw original image
timer_disp = deepcopy(orig_disp)
test_time = float(datetime.now().strftime('%s.%f')[9:-5])
if count != 0 and abs(test_time - curr_time) > 1:
curr_time = float(datetime.now().strftime('%s.%f')[9:-5])
count -= 1
timer_disp = timer(timer_disp, count)
if playing_surface:
# Configure and display the contoured and transformed images if the playing surface was found
cnt_disp = deepcopy(imutils.resize(playing_surface.img_cnt, height=300))
trans_disp = deepcopy(imutils.resize(playing_surface.transform, height=300))
display(timer_disp, cnt_disp, trans_disp)
valid_surface = playing_surface
else:
# Add a text overlay in place of the contour image
not_found_overlay = not_found(deepcopy(orig_disp))
if valid_surface is not None:
display(timer_disp, not_found_overlay, imutils.resize(
valid_surface.transform, height=300))
else:
# Display the countdown timer and the raw original until a valid surface is found
display(timer_disp, not_found_overlay)
# handling closing of displays in main atm, could shift anywhere though
key = cv2.waitKey(delay=1)
# keep displaying images until user enters 'q'
if key == ord('a') or key == ord('A'):
cv2.destroyAllWindows()
break
cv2.destroyAllWindows()
return valid_surface
def not_found(image):
text = 'Playing surface not found!'
font_pos = (10, image.shape[0] - 20)
# Give the font a little shadow to help it stand out
shadow_offset = 1
a = int(font_pos[0]) + shadow_offset
b = int(font_pos[1]) + shadow_offset
font_pos2 = (a, b)
font = cv2.FONT_HERSHEY_SIMPLEX
font_scale = 1
font_thickness = 2
line_type = cv2.LINE_AA
colour = (0, 0, 255)
# Add text to the image with a small shadow
cv2.putText(image, text, font_pos2, font, font_scale, (0, 0, 0), font_thickness, line_type)
cv2.putText(image, text, font_pos, font, font_scale, colour, font_thickness, line_type)
# Return the image with the text overlayed
return image