-
Notifications
You must be signed in to change notification settings - Fork 0
/
P5_get_topBarMsg.py
361 lines (310 loc) · 12.8 KB
/
P5_get_topBarMsg.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
"""
检查topBar中信息
"""
import copy
import os
import cv2
import os.path as osp
import numpy as np
from tqdm import tqdm
from P0_init import conf
from P2_img_split import get_image_names
import json
import re
import matplotlib.pyplot as plt
import math
from PaddleOCR.paddleocr import PaddleOCR
def init_rec_model(use_gpu,
rec_model_dir,
rec_char_dict_path,
det_model_dir,
cls_model_dir):
text_recognizer = PaddleOCR(det_model_dir=det_model_dir,
rec_model_dir=rec_model_dir,
cls_model_dir=cls_model_dir,
rec_char_dict_path=rec_char_dict_path,
use_gpu=use_gpu,
show_log=False)
return text_recognizer
def page_enhance(xin):
img_h, img_w = xin.shape[:2]
try:
img_c = xin.shape[2]
except IndexError:
img_c = 1
r = int(math.sqrt(float(img_w * img_h)) / 32.0) + 1
fin = cv2.GaussianBlur(xin, (r + r + 1, r + r + 1), 0.0)
xout = xin.copy()
for j in range(img_h):
po = xout[j]
pf = fin[j]
for i in range(img_w):
if img_c == 1:
xout[j][i] = 0 if pf[i] == 0 else min(
255.0, 255.0 * po[i] / pf[i])
else:
for ci in range(img_c):
xout[j][i][ci] = 0 if pf[i][ci] == 0 else min(
255.0, 255.0 * po[i][ci] / pf[i][ci])
return xout
rec_model_dir = 'PaddleOCR/inference/ch_ppocr_server_v2.0_rec_infer'
rec_char_dict_path = 'ppocr/utils/ppocr_keys_v1.txt'
det_model_dir = 'PaddleOCR/model/210924_ch_ppocr_server_v2.0_infer/det'
cls_model_dir = 'PaddleOCR/model/210924_ch_ppocr_server_v2.0_infer/cls'
use_gpu = conf['use gpu']
ocr_model = init_rec_model(use_gpu, rec_model_dir, rec_char_dict_path, det_model_dir, cls_model_dir)
kernel = np.ones((3, 3), np.uint8)
def get_topbar_data(baseRoot, specific_pages=None):
"""
得到顶栏信息(页码、批次)
:param image_root_dict:
:param specific_pages:
"""
image_root_dict = os.path.join(baseRoot, 'HP_roa_split/uppers')
img_names = get_image_names(image_root_dict, specific_pages)
page_infos = {}
for img_name in tqdm(img_names):
img_path = osp.join(image_root_dict, img_name)
ori_img = cv2.imread(img_path, 0)
page_number, class_info = check_page_data(ori_img)
tqdm.write('{}> page{},{}'.format(img_name, page_number, class_info))
page_infos[img_name] = {'page_number': page_number,
'class_info': class_info}
page_infos = recheck_page_number(page_infos)
page_infos = recheck_page_infos(page_infos)
json_page_msg = os.path.join(baseRoot, 'json/page_msg.json')
with open(json_page_msg, 'w', encoding='utf-8') as f:
f.write(json.dumps(page_infos, indent=4, ensure_ascii=False))
# with open(osp.join(image_root_dict, 'page_infos.json'), 'r', encoding='utf-8') as f:
# page_infos = json.load(f)
def recheck_page_infos(page_infos):
"""
重新对info dict进行整理,检查并订正错误的clas_info项
遍历每一页,若当前页的前后页具有相同info,则当前页也应具有相同info
:param page_infos:
:return: dict
"""
infos_per_page = {value['page_number']: value['class_info'] for key, value in page_infos.items()}
page_numbers = sorted(infos_per_page)
for p_idx in page_numbers:
if p_idx == -1:
continue
l_idx = p_idx - 1
r_idx = p_idx + 1
try:
if infos_per_page[l_idx] == infos_per_page[r_idx]:
infos_per_page[p_idx] = infos_per_page[l_idx]
except KeyError:
continue
new_infos = {}
for key, page_item in page_infos.items():
p_idx = page_item['page_number']
if p_idx == -1:
new_infos[key] = page_item
continue
info = page_item['class_info']
if infos_per_page[p_idx] != page_item['class_info']:
info = infos_per_page[p_idx]
new_infos[key] = dict(page_number=p_idx, class_info=info)
return new_infos
def recheck_page_number(page_infos):
"""
重新对info dict进行整理,检查并订正错误的page_number项
遍历每一页,若当前页的前后页page_number间隔1,则当前页的page_number为两者之间
!!使用前提是文件命名顺序正确!!
:param page_infos:
:return:
"""
img_names = list(page_infos.keys())
for idx, img_name in enumerate(img_names):
page_number = page_infos[img_name]['page_number']
if page_number != -1:
continue
img_num = int(img_name[:-4])
try:
l_img_num = int(img_names[idx - 1][:-4])
r_img_num = int(img_names[idx + 1][:-4])
l_info = page_infos[img_names[idx - 1]]['class_info']
r_info = page_infos[img_names[idx + 1]]['class_info']
l_page_number = page_infos[img_names[idx - 1]]['page_number']
r_page_number = page_infos[img_names[idx + 1]]['page_number']
except IndexError:
continue
if not (l_img_num + 1 == img_num == r_img_num - 1):
continue
if not (l_page_number + 2 == r_page_number):
continue
if l_info == r_info:
page_infos[img_name]['page_number'] = l_page_number + 1
page_infos[img_name]['class_info'] = l_info
return page_infos
def check_page_data(ori_img):
"""
分析传入图片并得到页码和批次信息
:param ori_img:
:return: int, str
"""
img_h, img_w = ori_img.shape[:2]
if img_h < 1000:
page_number, class_info = get_default_topbar_data(ori_img)
else:
page_number, class_info = get_title_topbar_data(ori_img)
return page_number, class_info
def get_default_topbar_data(ori_img):
"""
针对默认顶栏类型,获取其中信息(页码、批次)
:ori_img 裁剪后的topbar图片
:return: 页码, 批次
"""
ret, binary_img = cv2.threshold(ori_img, 200, 255, cv2.THRESH_BINARY)
left_bin_img = binary_img[:, :300]
right_bin_img = binary_img[:, -300:]
left_sum = sum(map(sum, left_bin_img))
right_sum = sum(map(sum, right_bin_img))
if left_sum > right_sum:
# 页码在右侧
page_number_img = ori_img[:, -300:]
class_info_img = ori_img[:, -3000:-300]
else:
# 页码在左侧
page_number_img = ori_img[:, :300]
class_info_img = ori_img[:, 300:3000]
page_number = get_default_pagenumber(page_number_img)
class_info = get_default_classinfo(class_info_img)
return page_number, class_info
def get_default_classinfo(class_info_img):
"""
针对默认顶栏中的批次信息图片,进行处理并ocr得到批次
:param class_info_img:
:return: 批次信息str
"""
class_info_img = page_enhance(class_info_img[:-20, :])
ret, class_info_img = cv2.threshold(class_info_img, 200, 255, cv2.THRESH_BINARY)
class_info_img = cv2.medianBlur(class_info_img, 3)
class_info_h, class_info_w = class_info_img.shape[:2]
# 每列求和
rows_sum = np.sum(class_info_img, axis=0)
rows_sum = list(map(lambda x: x / 255, rows_sum))
# 分别从前向后和从后向前找连续0的开始,即页码框部分 (0-黑 1-白)
flag_front, flag_behind = 0, 0
start_idx, end_idx = 0, 0
for idx in range(class_info_w):
flag_front = flag_front + 1 if rows_sum[idx] < (class_info_h - 10) and not start_idx else 0
flag_behind = flag_behind + 1 if rows_sum[-idx] < (class_info_h - 10) and not end_idx else 0
if flag_front >= 2:
start_idx = idx
if flag_behind > 2:
end_idx = idx
if start_idx and end_idx:
break
class_info_img = class_info_img[:, start_idx:class_info_w - end_idx]
ocr_results = ocr_model.ocr(class_info_img, det=False, cls=False)
ocr_results = sorted(ocr_results, key=lambda x: x[-1], reverse=True) # 按置信度降序排列
try:
result_text = ocr_results[0][0]
result_text = result_text[result_text.index('专刊') + 2:]
except IndexError:
result_text = ''
except ValueError:
print('{} not contains \"专刊\"!'.format(result_text))
class_info = result_text
# 利用正则匹配取出批次信息
# ^[,;.:.\"\'·]*([\u4e00-\u9fa5].*[\u4e00-\u9fa5])[,;.:.\"\'·]*$
# ^[^\u4e00-\u9fa5]*([\u4e00-\u9fa5].*[\u4e00-\u9fa5])[^0-9\u4e00-\u9fa5]*$
info_pattern = "^[^\u4e00-\u9fa5]*([\u4e00-\u9fa5].*[\u4e00-\u9fa5])[^0-9\u4e00-\u9fa5]*$"
class_pattern = '[\u4e00-\u9fa5].*[\u4e00-\u9fa5]'
if re.match(info_pattern, class_info) is not None:
class_info = re.search(class_pattern, class_info)[0]
return class_info
def get_default_pagenumber(page_number_img):
"""
针对默认顶栏中的页码图片,进行处理并ocr得到页码数字
:param page_number_img: 粗略裁剪的页码位置图片
:return: 页码数字
"""
# page_number_img = page_enhance(page_number_img[:-20, :])
ret, page_number_img = cv2.threshold(page_number_img[:-20, :], 195, 255, cv2.THRESH_BINARY)
# page_number_img = cv2.erode(page_number_img, np.ones((5, 5), np.uint8))
# page_number_img = cv2.medianBlur(page_number_img, 3)
page_number_h, page_number_w = page_number_img.shape[:2]
rows_sum = np.sum(page_number_img, axis=0)
rows_sum = list(map(lambda x: int(x / 255), rows_sum))
# 分别从前向后和从后向前找连续0的开始,即页码框部分 (0-黑 1-白)
flag_front, flag_behind = 0, 0
start_idx, end_idx = 0, 0
for idx in range(page_number_w):
flag_front = flag_front + 1 if rows_sum[idx] == 0 and not start_idx else 0
flag_behind = flag_behind + 1 if rows_sum[-idx] == 0 and not end_idx else 0
if flag_front >= 5 and start_idx == 0:
start_idx = idx
if flag_behind >= 5 and end_idx == 0:
end_idx = idx
if start_idx and end_idx:
break
page_number_img = page_number_img[:, start_idx:page_number_w - end_idx - 5]
page_number_img = ~page_number_img
page_number_img = cv2.erode(page_number_img, kernel=kernel, iterations=1)
page_number_img = cv2.medianBlur(page_number_img, 3)
plt.imshow(page_number_img, cmap=plt.get_cmap('gray'))
plt.axis('off')
plt.show()
ocr_results = ocr_model.ocr(page_number_img, det=False, cls=False)
try:
result = eval(ocr_results[0][0])
except IndexError:
result = -1
# print(result)
# cv2.imshow('pagenumber', page_number_img)
return result
def get_title_topbar_data(ori_img):
"""
针对封面标题顶栏类型,获取其中信息(页码、批次)
:param ori_img: 裁剪后的topbar图片
:return: 页码, 批次
"""
bar_top_pos = 985
bar_height = 100
bar_left_pos = 585
bar_width = 2500
bar_img = ori_img[bar_top_pos:bar_top_pos + bar_height, bar_left_pos:bar_left_pos + bar_width]
bar_img = page_enhance(bar_img)
bar_h, bar_w = bar_img.shape[:2]
ret, bar_img = cv2.threshold(bar_img, 200, 255, cv2.THRESH_BINARY)
bar_img = cv2.medianBlur(bar_img, 3)
# 每列求和
rows_sum = np.sum(bar_img, axis=0)
rows_sum = list(map(lambda x: x / 255, rows_sum))
flag_behind, end_idx = 0, 0
for idx in range(bar_w):
flag_behind = flag_behind + 1 if rows_sum[-idx] > bar_h * 2 else 0
if flag_behind >= 2:
end_idx = idx
break
bar_img = bar_img[:, :bar_w - end_idx]
ocr_results = ocr_model.ocr(bar_img, det=False, cls=False)
ocr_results = sorted(ocr_results, key=lambda x: x[-1], reverse=True) # 按置信度降序排列
try:
result_text = ocr_results[0][0]
result_text = result_text[result_text.index('专刊') + 2:]
except IndexError:
return -1, ''
except ValueError:
print('\"{}\" not contains \"专刊\"!'.format(result_text))
page_number = -1
class_info = result_text
# 正则表达式匹配,若符合则取出
# ^[,,;.:.\"\'·]*([\u4e00-\u9fa5].*[\u4e00-\u9fa5])[,,;.:.\"\'·]*[0-9]{1,3}[,;.:.\"\'·】\]》]*$
# ^[^\u4e00-\u9fa5]*([\u4e00-\u9fa5].*[\u4e00-\u9fa5])[^0-9\u4e00-\u9fa5]*[0-9]{1,3}[^0-9\u4e00-\u9fa5]*$
info_pattern = "^[^\u4e00-\u9fa5]*([\u4e00-\u9fa5].*[\u4e00-\u9fa5])" \
"[^0-9\u4e00-\u9fa5]*[0-9]{1,3}[^0-9\u4e00-\u9fa5]*$"
number_pattern = '[0-9]{1,3}'
class_pattern = '[\u4e00-\u9fa5].*[\u4e00-\u9fa5]'
if re.match(info_pattern, result_text) is not None:
page_number = eval(re.search(number_pattern, result_text)[0])
class_info = re.search(class_pattern, result_text)[0]
return page_number, class_info
def main():
# get_topbar_data('PC2/HP_upper/', list(range(0, 2)))
get_topbar_data("PC_art_all")
if __name__ == "__main__":
main()