-
Notifications
You must be signed in to change notification settings - Fork 1
/
segmentar.py
523 lines (447 loc) · 19.5 KB
/
segmentar.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
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
"""
This script contains all the codes for segmenting the Indian number plates
It contains a hierarchial procedure for extracting very noisy and bad images in Indian number plates
"""
import numpy as np
import cv2
import imutils
import matplotlib.pyplot as plt
import statsmodels.api as sm
from scipy.signal import argrelextrema
W=320 # width is fixed
def fill_dirn(thresh_img, dirn):
"""
It is used for filling
Give this function an image and direction, it will check in that direction and if encounters a white pixel it will make all other remaining pixels in that direction white.
Arguments:
dirn: direction of filling - includes
ht - horizontal top
hb - horizontal bottom
vl - vertical left
vr - vertical right
Returns:
Filled image along a particular direction
"""
thresh_x = thresh_img.copy()
if dirn == "ht":
for dirn_i in range(thresh_img.shape[1]):
for dirn_j in range(thresh_img.shape[0]):
if thresh_img[dirn_j][dirn_i]==255:
thresh_x[dirn_j:, dirn_i]=255
break
if dirn == "hb":
for dirn_i in range(thresh_img.shape[1]):
for dirn_j in range(thresh_img.shape[0]):
if thresh_img[thresh_img.shape[0]-1-dirn_j][dirn_i]==255:
thresh_x[0:thresh_img.shape[0]-dirn_j, dirn_i]=255
break
if dirn == "vl":
for dirn_i in range(thresh_img.shape[0]):
for dirn_j in range(thresh_img.shape[1]):
if thresh_img[dirn_i][dirn_j]==255:
thresh_x[dirn_i, dirn_j:]=255
break
if dirn == "vr":
for dirn_i in range(thresh_img.shape[0]):
for dirn_j in range(thresh_img.shape[1]):
if thresh_img[dirn_i][thresh_img.shape[1]-1-dirn_j]==255:
thresh_x[dirn_i, 0:thresh_img.shape[1]-dirn_j]=255
break
return thresh_x.astype("uint8")
def smooth_data_convolve_my_average(arr, span): ## Function for smoothing the curve
re = np.convolve(arr, np.ones(span * 2 + 1) / (span * 2 + 1), mode="same")
# The "my_average" part: shrinks the averaging window on the side that
# reaches beyond the data, keeps the other side the same size as given
# by "span"
re[0] = np.average(arr[:span])
for i in range(1, span + 1):
re[i] = np.average(arr[:i + span])
re[-i] = np.average(arr[-i - span:])
return re
def v_projection_or_bruteforce_trimming(img, half_detected = True, min=None): # Our novel proposed and implemeted approach for bigger noiser data
# But this approach have some false positive which can be addressed
# in future if we get more time,
"""
This function is used for divind a giving image into subparts if multiple letters are present else return the same image
"""
h, w = img.shape
combined_list = [0] # list contains max and intial and final
split_imgs = []
#vertical_projection
col_pix_count = np.sum(img, axis = 0)
# Using moving average method with scipy lowess to find the maxima and reduce noise to cut the parts,-----------
maxs = argrelextrema(smooth_data_convolve_my_average(col_pix_count, 3), np.less)[0]
combined_list.extend(maxs)
combined_list.append(w-1)
# print(combined_list)
if len(maxs)==0:
return [img]
elif half_detected: # if segmentation is half done already then
filtered_list = [0]
i=0
j = 0
# print("Minimum : ",min)
if combined_list[-1]<=min:
return [img]
if combined_list[-1]>40:
min = 17
while((i < len(combined_list)-1) and (i+j < len(combined_list)-1)):
diff = -(combined_list[i]-combined_list[i+1+j])
if diff<=min:
j+=1
else:
if -(combined_list[i+j+1]-combined_list[-1])<min:
filtered_list.append(combined_list[-1])
break
filtered_list.append(combined_list[i+j+1])
i+=j+1
j=0
else:
filtered_list = [0]
i=0
j = 0
min=10
while((i < len(combined_list)-1) and (i+j < len(combined_list)-1)):
diff = -(combined_list[i]-combined_list[i+1+j])
if diff<=min:
j+=1
else:
if -(combined_list[i+j+1]-combined_list[-1])<min:
filtered_list.append(combined_list[-1])
break
filtered_list.append(combined_list[i+j+1])
i+=j+1
j=0
# print(filtered_list)
#plt.plot(col_pix_count, color = "b")
#plt.plot(smooth_data_convolve_my_average(col_pix_count, 3), color = "r")
#plt.show()
for i in range(len(filtered_list)-1):
split_imgs.append(img[:,filtered_list[i]: filtered_list[i+1]])
return split_imgs
def extract_plate(gray):
"""
this function provides the approx polyDP coordinates which includes the plate.
If it can't detect the plate and can't fulfil the area constraint, it will return the approx of whole image
Arguments:
gray: Grayscale image
Retuens:
approx points along with binary thresholded image
"""
h1, w1 = gray.shape
dilated_img = cv2.dilate(gray, np.ones((7, 7), np.uint8))
bg_img = cv2.bilateralFilter(dilated_img, 11, 17, 17) #using median blur to remove the undesired shadow along with abs difference and normalization
diff_img = 255 - cv2.absdiff(gray, bg_img)
norm_img = cv2.normalize(diff_img,None, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_8UC1)
ret,binary1 = cv2.threshold(norm_img, 100, 255, cv2.THRESH_OTSU | cv2.THRESH_BINARY)
# binary = check_h_border(binary)
# binary1 = check_v_border(binary)
binary = cv2.bitwise_not(binary1)
# cv2.imshow("BINARY", binary)
""" NOW we will use the vertical filling to determine the effective area of number plate """
### ----------------------------------------Firstly calculating and using vertival sobel---------------------------------#####
V = cv2.Sobel(binary, cv2.CV_8U, 2, 0)
V= cv2.dilate(V, np.ones((3,2), np.uint8), iterations =1)
contours,_ = cv2.findContours(V, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
for cnt in contours:
(x,y,w,h) = cv2.boundingRect(cnt)
# rows/3 is the threshold for length of line
if h > h1/4:
cv2.drawContours(V, [cnt], -1, 255, -1)
cv2.drawContours(binary, [cnt], -1, 255, -1)
else:
cv2.drawContours(V, [cnt], -1, 0, -1)
# cv2.imshow("Sobel", V)
kernel = cv2.getStructuringElement(shape=cv2.MORPH_RECT, ksize=(3,3))
V = cv2.morphologyEx(V, cv2.MORPH_DILATE, kernel, iterations = 3)
thresh_vr = fill_dirn(V, "vr")
thresh_vl = fill_dirn(V, "vl")
thresh_v = cv2.bitwise_and(thresh_vl, thresh_vr)
#cv2.imshow("THRESH_V", thresh_v)
nnx = np.zeros(thresh_v.shape, np.uint8)
cnts = cv2.findContours(thresh_v, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
cnts = imutils.grab_contours(cnts)
c = sorted(cnts, key=cv2.contourArea, reverse=True)
if len(c)==0:
approx_f = np.array([[[0, 0]], [[w1, 0]], [[w1, h1]], [[0, h1]]])
return approx_f, binary1
else:
c = c[0]
coordi2 = cv2.boundingRect(c)
width_v = coordi2[2]
hull = cv2.convexHull(c, False)
cv2.drawContours(nnx, [hull], 0, 255, -1, 8)
cv2.drawContours(nnx, [hull], 0, 255, 5, 8)# after getting the mask extending its boundary to get a bigger area
# cv2.imshow("NNX", nnx)
###----------------For H-SOBEL----------------------------------------
H = cv2.Sobel(binary, cv2.CV_8U, 0, 2)
H= cv2.dilate(H, np.ones((2,3), np.uint8), iterations =1)
thresh_ht = fill_dirn(H, "hb")
thresh_hb = fill_dirn(H, "ht")
thresh_h = cv2.bitwise_and(thresh_ht, thresh_hb)
cnts = cv2.findContours(thresh_h, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
cnts = imutils.grab_contours(cnts)
cnts = sorted(cnts, key = cv2.contourArea, reverse =True)[0]
coordi = cv2.boundingRect(cnts)
width_h = coordi[2]
cv2.rectangle(thresh_h, (coordi[0], coordi[1]), (coordi[0]+coordi[2], coordi[1]+coordi[3]), 255, -1)
# cv2.imshow("Sobel-H", H)
# cv2.imshow("thresh_h", thresh_h)
##-------------------------------------------------------------------------
#####------------------Taking help of both V and H sobel to get the resultant nxx mask----------------
# print(width_v/width_h)
if width_v/width_h < 0.6: ## if width relativeness less than 0.6 so detect bigger width
cv2.rectangle(nnx, (coordi[0], coordi2[1]), (coordi2[0]+coordi2[2], coordi2[1]+coordi2[3]), 255, -1)
# cv2.imshow("nnx", nnx)
nnx_dst = nnx.copy()
cnts, _ = cv2.findContours(nnx, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
c = sorted(cnts, key=cv2.contourArea, reverse=True)
if len(c)==0:
approx_f = np.array([[[0, 0]], [[w1, 0]], [[w1, h1]], [[0, h1]]])
return approx_f, binary1
else:
c = c[0]
approx_f = cv2.approxPolyDP(c, 0.05*cv2.arcLength(c, True), True)# Tested on different images 0.06 is suitable for most of them
# print("Approx_f :",len(approx_f))
if len(approx_f)==4:
# If rectangle detected
print("RECT DETECTED SUCCESSFULLY")
else:
# If mask detected but rect not detected returning bounding rect of mask (can be changed to minAreaRect)
re = cv2.boundingRect(c)
cv2.rectangle(nnx_dst, (re[0], re[1]), (re[0]+re[2], re[1]+re[3]), 255, -1)
cnts, _ = cv2.findContours(nnx_dst, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
c = sorted(cnts, key=cv2.contourArea, reverse=True)[0]
approx_f = cv2.approxPolyDP(c, 0.05*cv2.arcLength(c, True), True)
if len(approx_f!=4):
approx_f = np.array([[[0, 0]], [[w1, 0]], [[w1, h1]], [[0, h1]]])
if cv2.contourArea(c)<(0.2*w1*h1):
# If mask not detected returns approx of whole image
approx_f = np.array([[[0, 0]], [[w1, 0]], [[w1, h1]], [[0, h1]]])
return approx_f, binary1
def check_h_border(thresh_img):
n_row, n_column = thresh_img.shape
thresh_img = cv2.bitwise_not(thresh_img)
prs = 0
i = 0
while(i<(n_row//5)):
row = thresh_img[i]
row = row/255
rs = np.sum(row)
if rs<(prs//2):
thresh_img[:i] = 0
else:
if rs>=prs:
prs = rs
i += 5
thresh_img = cv2.flip(thresh_img, 0)
prs = 0
i = 0
while(i<(n_row//5)):
row = thresh_img[i]
row = row/255
rs = np.sum(row)
if rs<(prs//2):
thresh_img[:i] = 0
else:
if rs>=prs:
prs = rs
i += 5
thresh_img = cv2.flip(thresh_img, 0)
thresh_img = cv2.bitwise_not(thresh_img)
return thresh_img
def check_v_border(thresh_img):
n_row, n_column = thresh_img.shape
thresh_img = cv2.bitwise_not(thresh_img)
pcs = 0
i = 0
while(i<(n_column//15)):
column = thresh_img[:, i]
column = column/255
cs = np.sum(column)
if cs<(pcs//2):
thresh_img[:, :i] = 0
break
else:
if cs>=pcs:
pcs = cs
i += 5
thresh_img = cv2.flip(thresh_img, 1)
pcs = 0
i = 0
while(i<(n_column//15)):
column = thresh_img[:, i]
column = column/255
cs = np.sum(column)
if cs<(pcs//2):
thresh_img[:, :i] = 0
break
else:
if cs>=pcs:
pcs = cs
i += 5
thresh_img = cv2.flip(thresh_img, 1)
thresh_img = cv2.bitwise_not(thresh_img)
return thresh_img
def order_points(pts):
# initialzie a list of coordinates that will be ordered
# such that the first entry in the list is the top-left,
# the second entry is the top-right, the third is the
# bottom-right, and the fourth is the bottom-left
rect = np.zeros((4, 2), dtype = "float32")
# the top-left point will have the smallest sum, whereas
# the bottom-right point will have the largest sum
s = pts.sum(axis = 1)
rect[0] = pts[np.argmin(s)]
rect[2] = pts[np.argmax(s)]
# now, compute the difference between the points, the
# top-right point will have the smallest difference,
# whereas the bottom-left will have the largest difference
diff = np.diff(pts, axis = 1)
rect[1] = pts[np.argmin(diff)]
rect[3] = pts[np.argmax(diff)]
# return the ordered coordinates
return rect
def four_point_transform(image, pts):
# obtain a consistent order of the points and unpack them
# individually
pts= pts.reshape(4,2)
rect = order_points(pts)
(tl, tr, br, bl) = rect
# compute the width of the new image, which will be the
# maximum distance between bottom-right and bottom-left
# x-coordiates or the top-right and top-left x-coordinates
widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
maxWidth = max(int(widthA), int(widthB))
# compute the height of the new image, which will be the
# maximum distance between the top-right and bottom-right
# y-coordinates or the top-left and bottom-left y-coordinates
heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
maxHeight = max(int(heightA), int(heightB))
# now that we have the dimensions of the new image, construct
# the set of destination points to obtain a "birds eye view",
# (i.e. top-down view) of the image, again specifying points
# in the top-left, top-right, bottom-right, and bottom-left
# order
dst = np.array([
[0, 0],
[maxWidth - 1, 0],
[maxWidth - 1, maxHeight - 1],
[0, maxHeight - 1]], dtype = "float32")
# compute the perspective transform matrix and then apply it
M = cv2.getPerspectiveTransform(rect, dst)
warped = cv2.warpPerspective(image, M, (maxWidth-1, maxHeight-1), flags = cv2.INTER_AREA)
# return the warped image
return warped
def sort_x(cnt):
"""
Retuening the width value of a contour for sorting
"""
return cnt[0]
def extraction(path):
"""
Final generator type function for detecting the plate and combining all helper functions to segment the letters
"""
image = cv2.imread(path, cv2.IMREAD_UNCHANGED)
h, w = image.shape[:2]
H = int(W*h/w)
image = cv2.resize(image, (W, H), cv2.INTER_AREA)
image = cv2.normalize(image, image, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX)
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# Plate Exatraction
aprx_f, binary = extract_plate(gray.copy())
nnx_f = np.zeros(binary.shape, np.uint8)
nnx_f = cv2.drawContours(nnx_f, [aprx_f], 0, 255, -1)
# cv2.imshow("NNX_F", nnx_f)
wraped = cv2.bitwise_and(binary, binary, mask = nnx_f)
# Four point Transform
wraped = four_point_transform(wraped, aprx_f)
# bg_img = cv2.bilateralFilter(wraped, 13, 15, 15) #using median blur to remove the undesired shadow along with abs difference and normalization
# _, thresh = cv2.threshold(bg_img, 110, 255, cv2.THRESH_BINARY|cv2.THRESH_OTSU)
thresh_n = wraped.copy()
# cv2.imshow("wraped", wraped)
thresh_n = cv2.copyMakeBorder(thresh_n, 5, 5, 5, 5, cv2.BORDER_CONSTANT, value=255)
cnts, _ = cv2.findContours(thresh_n, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)
thresh_nn = np.zeros(thresh_n.shape, np.uint8)
rects = []
for i, c in enumerate(cnts):
rect_t = cv2.boundingRect(c)
if ((rect_t[2]*rect_t[3])<(0.7*w*h) and (rect_t[2]*rect_t[3])>160) and rect_t[2]>10:
thresh_nn = cv2.drawContours(thresh_nn, [c], 0, 255, -1)
# print(rect_t[2]*rect_t[3])
# From the previous black region detecting bonding rects and cropping that portion from the previous threshold image (thresh_n)
# This time tere will be only one contour for a character at max because its inside region is filled
#thresh_nn = cv2.erode(thresh_nn, np.ones((5, 5), np.uint8), iterations=1)
cnts, _ = cv2.findContours(thresh_nn, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)
rects = []
print("Contours : ",len(cnts))
for c in cnts:
rect_t = cv2.boundingRect(c)
if ((rect_t[2]*rect_t[3])<(0.7*w*h) and (rect_t[2]*rect_t[3])>160) and (rect_t[2]>10):
# print(rect_t[2]*rect_t[3])
rects.append(rect_t)
#-------------------Now return the characters if segmented else applying further cutting -----------#
chars = []
if len(rects)!=0: # means at least some characters detected
rects = sorted(rects, key = sort_x)
#Appraoch-2 here starts the second hierarchial approach after normal sobel technique
widths = np.array([rect[2] for rect in rects])
median = np.median(widths[np.argsort(widths)])
# Sowing the extracted rects
for ind, rect_t in enumerate(rects):
# Get the 4 points of the bounding rectangle
x, y, w, h = rect_t
# Draw a straight rectangle with the points
# cv2.imshow("THRESH_NN", thresh_nn)
s = thresh_n[y:(y+h), x:(x+w)]
s = cv2.bitwise_not(s)
trimmed = v_projection_or_bruteforce_trimming(s, half_detected = True, min = median)
for trim in trimmed:
h1, w1 = trim.shape
if h1>w1:
diff = h1-w1
trim = cv2.copyMakeBorder(trim, 4, 4, 2, 2, cv2.BORDER_CONSTANT, value=0)
else:
diff = w1-h1
trim = cv2.copyMakeBorder(trim, 2, 2, 4, 4, cv2.BORDER_CONSTANT, value=0)
trim = cv2.resize(trim, (64,64), cv2.INTER_AREA)
#trim = cv2.erode(trim, np.ones((2,2), np.uint8), iterations = 1)
chars.append(trim)
# cv2.imshow("final", cv2.bitwise_not(cv2.bitwise_and(binary, binary, mask = nnx_f)))
else:
## -----Here we will use the concept of minimum extraction to reove the plates--------------##
h,w = thresh_n.shape
final = cv2.bitwise_not(cv2.bitwise_and(binary, binary, mask = nnx_f))
final = four_point_transform(final, aprx_f)
# cv2.imshow("final", final)
trimmed = v_projection_or_bruteforce_trimming(final, half_detected = False)
for trim in trimmed:
h1, w1 = trim.shape
if h1>w1:
diff = h1-w1
trim = cv2.copyMakeBorder(trim, 4, 4, 2, 2, cv2.BORDER_CONSTANT, value=0)
else:
diff = w1-h1
trim = cv2.copyMakeBorder(trim, 2, 2, 4, 4, cv2.BORDER_CONSTANT, value=0)
trim = cv2.resize(trim, (64,64), cv2.INTER_AREA)
# trim = cv2.erode(trim, np.ones((2,2), np.uint8), iterations = 1)
chars.append(trim)
# cv2.imshow("original", image)
return chars
if __name__ == '__main__':
extraction(input("Path: \n"))
cv2.waitKey(0)
cv2.destroyAllWindows()
"""
SP(1 L)
Y
Y
SP
SP
Y
SP
"""