最近做了个手写汉字简历识别比赛，需要先提取表格中含有指定信息的各个框，再用TensorFlow对框中的信息进行汉字、数字、英文识别。代码已开源：https://github.com/BingLiHanShuang/chinese_ocr，需要额外下载训练好的模型文件：https://pan.baidu.com/s/1Q0dPSKILNxPMDn7i2VIhow （或不使用百度云（拷进网址栏）：https://d.pcs.baidu.com/file/8c755e26ce17286ed911405a7644f73f?fid=4031771572-250528-760858459813394&dstime=1544602011&rt=sh&sign=FDtAERVY-DCb740ccc5511e5e8fedcff06b081203-GHzypluGZoBNdm99UChQOUdYJJ0%3D&expires=8h&chkv=1&chkbd=0&chkpc=et&dp-logid=8017096450883646029&dp-callid=0&shareid=3190055425&r=806607689 ）
整体的工作量有点大，此篇先介绍如何对表格中各个框进行提取，此部分代码位于table_choose.py文件中的前半部分，本文中摘取部分进行说明，具体可见上文开源Github代码链接。笔者先前用的都是C++版的OpenCV，因为TensorFlow识别部分的代码基于python，为了减少工作量，这次就顺便使用了Python版的OpenCV，需要在OpenCV编译时补全Python项，并复制cv2.so和cv.py文件【重要！否则import cv2会提示找不到module，见https://www.cnblogs.com/freeweb/p/5794447.html】。
如今OpenCV官方的Document里都包含了各API的Python版用法，从C++转换到Python用起来意外的顺手。这里附上官方的在线Document地址，记得选对应版本：https://docs.opencv.org/
笔者该部分的编译环境如下：
（1）Ubuntu16.04 64位 支持utf-8编码
（2）Python2.7
（3）OpenCV3.4.3 （与Python2.7编译通）
（4）Python2.7 numpy模块、PIL模块、logging模块、pickle模块、os模块、random模块、time模块、matplotlib模块、math模块、csv模块
先一睹Python+OpenCV校正并提取表格中的各个框的整体代码table_choose.py（其中包含识别部分的功能，提取表格的功能需要摘出来使用）：
#!/usr/bin/env python
# encoding: utf-8
import os
import cv2 as cv
import numpy as np
import math
import num_out
import mnist_recognize
import chinese_out
import chinese_out_jiguan
import time_out
import csv
##### 初始化CSV文件 #####
csvfile = open("try1010.csv", 'w')
csvwrite = csv.writer(csvfile)
fileHeader = ["登记表编号","性别","民族","体重","血型","籍贯","高中学校名称","高中专业","高中学位","高中起止时间","高中是否毕业","大专学校名称","大专专业","大专学位","大专起止时间","大专是否毕业","本科学校名称","本科专业","本科学位","本科起止时间","本科是否毕业","研究生学校名称","研究生专业","研究生学位","研究生起止时间","研究生是否毕业","学校名称（其它）","专业（其它）","学位（其它）","起止时间（其它）","是否毕业（其它）"]
csvwrite.writerow(fileHeader)
path = 'test_data'
for (path,dirs,files) in os.walk(path):
for filename in files:
#docu_num = '20110158' #测试单张登记表
##### 按文件名读取登记表 #####
(docu_num,extension) = os.path.splitext(filename)
print "filename:", filename
##### 根据文件名年份的不同区分不同类型的登记表 #####
print docu_num[3]
if docu_num[3] == '1':
if len(docu_num) == 9 or (docu_num[5] == '3' and docu_num[6] == '1'):
table_style = 2
else:
table_style = 1
if docu_num[3] == '3':
table_style = 3
if docu_num[3] == '4' or docu_num[3] == '5':
table_style = 4
jiaozheng = 0
image = cv.imread('test_data/' + docu_num + '.jpg')
rows, cols, channels = image.shape
print rows, cols
image_copy = image.copy()
##### 旋转校正Rotation #####
#统计图中长横线的斜率来判断整体需要旋转矫正的角度
gray = cv.cvtColor(image, cv.COLOR_RGB2GRAY)
if table_style == 1 or table_style == 3 or table_style == 2:
edges = cv.Canny(gray, 50, 150, apertureSize=3) # 50,150,3
cv.imwrite('edges_whole.jpg', edges)
lines = cv.HoughLinesP(edges, 1, np.pi / 180, 500, 0, minLineLength=50, maxLineGap=50)#650,50,20
if table_style == 4:
edges_gray = cv.Canny(gray, 50, 150, apertureSize=3) # 50,150,3
edges = edges_gray[400:1000, 0:1000]
cv.imwrite('edges_whole.jpg', edges)
lines = cv.HoughLinesP(edges, 1, np.pi / 180, 200, 0, minLineLength=50, maxLineGap=35)#650,50,20
pi = 3.1415
theta_total = 0
theta_count = 0
for line in lines:
x1, y1, x2, y2 = line[0]
if table_style == 4:
y1 = y1 + 400
y2 = y2 + 400
rho = np.sqrt((x1 - x2)**2 + (y1 - y2)**2)
theta = math.atan(float(y2 - y1)/float(x2 - x1 + 0.001))
print(rho, theta, x1, y1, x2, y2)
if theta < pi/4 and theta > -pi/4:
theta_total = theta_total + theta
theta_count+=1
cv.line(image_copy, (x1, y1), (x2, y2), (0, 0, 255), 2)
#cv.line(edges, (x1, y1), (x2, y2), (0, 0, 0), 2)
theta_average = theta_total/theta_count
print theta_average, theta_average*180/pi
cv.imwrite('line_detect4rotation.jpg', image_copy)
#cv.imwrite('line_detect4rotation.jpg', ~edges)
#affineShrinkTranslationRotation = cv.getRotationMatrix2D((cols/2, rows/2), theta_average*180/pi, 1)
affineShrinkTranslationRotation = cv.getRotationMatrix2D((0, rows), theta_average*180/pi, 1)
ShrinkTranslationRotation = cv.warpAffine(image, affineShrinkTranslationRotation, (cols, rows))
image_copy = cv.warpAffine(image_copy, affineShrinkTranslationRotation, (cols, rows))
cv.imwrite('image_Rotation.jpg',ShrinkTranslationRotation)
##### 平移校正Move #####
#通过对表格左下角直角进行识别，将其顶点统一平移矫正至(78,1581)
#print "rows: ",rows
roi = image_copy[1450:rows, 0:150]#180
gray = cv.cvtColor(roi, cv.COLOR_RGB2GRAY)
edges = cv.Canny(gray, 50, 150, apertureSize=3) # 50,150,3
roi_mean_set = cv.mean(~edges[0:int((rows-1450)/2), 85:150])#通过区域灰度值特征排除文字对直线识别的干扰
roi_mean = roi_mean_set[0]
#cv.imwrite('edges_sample.jpg', edges)
cv.imwrite('edges_sample.jpg', ~edges[0:int((rows-1450)/2), 75:150])
lines = cv.HoughLinesP(edges, 1.0, np.pi / 180, 35, 0, minLineLength=10,maxLineGap=20)#50,10,20
lines_message_set = []
for line in lines:
x1, y1, x2, y2 = line[0]
y1 = y1 + 1450
y2 = y2 + 1450
rho = np.sqrt((x1 - x2)**2 + (y1 - y2)**2)
xielv = (y2 - y1)/(x2 - x1 + 0.001)
theta = math.atan(float(y2 - y1)/float(x2 - x1 + 0.001))
print(rho, theta, x1, y1, x2, y2, xielv)
lines_message = (rho, theta, x1, y1, x2, y2, xielv)
#cv.line(image, (x1, y1), (x2, y2), (0, 0, 255), 2)
lines_message_set.append(lines_message)
#print len(lines_message_set)
#求点到直线的距离
def point2line_distance(x1,y1,x2,y2,pointPx,pointPy):
A = y1 - y2
B = x2 - x1
C = x1*y2 - y1*x2
distance = abs(A*pointPx + B*pointPy + C)/((A*A + B*B)**0.5)
return distance
lines_cluster_set = []
repeat_num_set = []
for j in range(0,len(lines_message_set)):
for i in range(0,len(lines_message_set)):
if not(j in repeat_num_set):
lines_cluster_set.append(lines_message_set[j])
repeat_num_set.append(j)
print point2line_distance(lines_message_set[i][2], lines_message_set[i][3], lines_message_set[i][4], lines_message_set[i][5], (lines_message_set[j][2]+lines_message_set[j][4])/2, (lines_message_set[j][3]+lines_message_set[j][5])/2)
if i!=j and abs(lines_message_set[j][6] - lines_message_set[i][6]) < 0.1 and point2line_distance(lines_message_set[i][2], lines_message_set[i][3], lines_message_set[i][4], lines_message_set[i][5], (lines_message_set[j][2]+lines_message_set[j][4])/2, (lines_message_set[j][3]+lines_message_set[j][5])/2) <= 10:
repeat_num_set.append(i)
print lines_cluster_set
#对直角的横线、竖线进行分析，缺省时根据表格类型进行矫正
Point_heng = []
Point_shu = []
MiddlePoint_heng = (112,1450+(rows-1450)/2)
MiddlePoint_shu = (75,1450+(rows-1450)/4)
distance2point = rows-1450
distance2point2 = rows-1450
for j in range(0,len(lines_cluster_set)):
if abs(lines_cluster_set[j][6]) < 1:
if ((MiddlePoint_heng[0]-(lines_cluster_set[j][2]+lines_cluster_set[j][4])/2)**2 + (MiddlePoint_heng[1]-(lines_cluster_set[j][3]+lines_cluster_set[j][5])/2)**2)**0.5 < distance2point:
distance2point = ((MiddlePoint_heng[0]-(lines_cluster_set[j][2]+lines_cluster_set[j][4])/2)**2 + (MiddlePoint_heng[1]-(lines_cluster_set[j][3]+lines_cluster_set[j][5])/2)**2)**0.5
Point_heng = lines_cluster_set[j]
else:
if ((MiddlePoint_shu[0]-(lines_cluster_set[j][2]+lines_cluster_set[j][4])/2)**2 + (MiddlePoint_shu[1]-(lines_cluster_set[j][3]+lines_cluster_set[j][5])/2)**2)**0.5 < distance2point2:
distance2point2 = ((MiddlePoint_shu[0]-(lines_cluster_set[j][2]+lines_cluster_set[j][4])/2)**2 + (MiddlePoint_shu[1]-(lines_cluster_set[j][3]+lines_cluster_set[j][5])/2)**2)**0.5
Point_shu = lines_cluster_set[j]
need_stronger = 0
something_missing = 0
#缺省矫正
if Point_shu != []:
if Point_heng == []:
cross_x = 78
cross_y = 1616
something_missing = 1
if len(docu_num) == 9:
something_missing = 0
cross_x = 93
cross_y = 1666
if docu_num[3] == '4':
something_missing = 0
cross_x = 78
cross_y = 1655
if docu_num[3] == '4' and docu_num[5] == '6':
cross_x = 78
cross_y = 1665
else:
cross_x = (Point_shu[3]-Point_shu[6]*Point_shu[2]-Point_heng[3]+Point_heng[6]*Point_heng[2])/(Point_heng[6]-Point_shu[6])
cross_y = Point_heng[6]*cross_x + Point_heng[3] - Point_heng[6]*Point_heng[2]
cv.line(image_copy, (Point_heng[2], Point_heng[3]), (Point_heng[4], Point_heng[5]), (0, 0, 255), 2)
cv.line(image_copy, (Point_shu[2], Point_shu[3]), (Point_shu[4], Point_shu[5]), (0, 0, 255), 2)
else:
cross_x = Point_heng[2]
cross_y = Point_heng[3]
need_stronger = 1
if Point_heng != [] and cross_x > Point_heng[2] and len(docu_num) == 9 and docu_num[6] == '0' and docu_num[7] == '9':
cross_x = 50
cross_y = 1586
if Point_heng != [] and cross_x > Point_heng[2] and len(docu_num) == 9 and docu_num[6] == '1' and docu_num[7] == '8':
cross_x = 50
cross_y = 1616
print 'roi_mean:',roi_mean
if len(docu_num) == 9 and docu_num[7] == '8' and roi_mean < 230:
cross_x = 78
cross_y = 1631
if cross_y - 1648 < 3:
if len(docu_num) == 8 and docu_num[5] == '2' and docu_num[6] == '4':
cross_x = 78
cross_y = 1601
if len(docu_num) == 9 and docu_num[6] == '1' and docu_num[7] == '4':
cross_x = 78
cross_y = 1621
if table_style == 3 and cross_y < 1485:
cross_x = 78
cross_y = 1641
print cross_x,cross_y#当下直角顶点位置，标准位置为78,1581
cv.circle(image_copy, (int(cross_x), int(cross_y)), 3,(255,0,0),3)
cv.rectangle(image_copy,(0,1450),(180,rows),(255,0,0),3)
cv.imwrite('line_detect_possible_demo.jpg', image_copy)
rows, cols, channels = ShrinkTranslationRotation.shape
print rows, cols
affineShrinkTranslation = np.array([[1, 0, int(78 - cross_x)], [0, 1, int(1581 - cross_y)]], np.float32)
#affineShrinkTranslation = np.array([[1, 0, int(78 - 78)], [0, 1, int(1581 - 1581)]], np.float32)
shrinkTwoTimesTranslation = cv.warpAffine(ShrinkTranslationRotation, affineShrinkTranslation, (cols, rows))
image_copy = cv.warpAffine(image_copy, affineShrinkTranslation, (cols, rows))
##### 对201100XXX中左下角表格内有额外竖线的表格进行检测Detect_not_shu_line_in_201100XXX #####
if table_style == 2:
shrinkTwoTimesTranslation_copy_copy = shrinkTwoTimesTranslation.copy()
roi = shrinkTwoTimesTranslation[1350:1548, 125:300]#180
cv.rectangle(image_copy,(125,1350),(300,1548),(255,0,0),3)
gray = cv.cvtColor(roi, cv.COLOR_RGB2GRAY)
edges = cv.Canny(gray, 50, 150, apertureSize=3) # 50,150,3
lines = []
lines = cv.HoughLinesP(edges, 1.0, np.pi / 180, 100, 0, minLineLength=60,maxLineGap=20)#50,10,20
print lines
go_through = 0
try:
if lines == None:
go_through = 0
except:
go_through = 1
if go_through == 1:
pi = 3.1415
theta_total = 0
theta_count = 0
for line in lines:
x1, y1, x2, y2 = line[0]
x1 = x1 + 125
x2 = x2 + 125
y1 = y1 + 1350
y2 = y2 + 1350
rho = np.sqrt((x1 - x2)**2 + (y1 - y2)**2)
theta = math.atan(float(y2 - y1)/float(x2 - x1 + 0.001))
print(rho, theta, x1, y1, x2, y2)
if theta > pi/3 or theta < -pi/3:
table_style = 1
if len(docu_num) == 9 and docu_num[6] == '1' and docu_num[7] == '1':
jiaozheng = 1
cv.line(shrinkTwoTimesTranslation_copy_copy, (x1, y1), (x2, y2), (255, 0, 0), 2)
cv.imwrite('shrinkTwoTimesTranslation_copy_copy.jpg', shrinkTwoTimesTranslation_copy_copy)
##### 提取表格Table_Out #####
#分别通过对二值化后的表格用长横条、长竖条内核进行开操作，将表格分别化为全横线与全竖线，叠加后提取交点，即可得到表格中每个矩形的四个顶点
shrinkTwoTimesTranslation_gray = cv.cvtColor(shrinkTwoTimesTranslation, cv.COLOR_RGB2GRAY)
th2 = cv.adaptiveThreshold(~shrinkTwoTimesTranslation_gray,255,cv.ADAPTIVE_THRESH_MEAN_C,cv.THRESH_BINARY,15,-2)
cv.imwrite('th2.jpg', th2)
#长横条内核处理
shrinkTwoTimesTranslation_copy = shrinkTwoTimesTranslation.copy()
th2_copy = th2.copy()
scale = 44;#20,50,45,40,44
rows,cols = shrinkTwoTimesTranslation_gray.shape
horizontalsize = cols / scale
horizontalStructure = cv.getStructuringElement(cv.MORPH_RECT, (horizontalsize, 1))
erosion = cv.erode(th2,horizontalStructure,iterations = 1)
dilation = cv.dilate(erosion,horizontalStructure,iterations = 1)
#长竖条内核处理
scale = 39;#20,50,45,40,39
horizontalsize2 = rows / scale
horizontalStructure2 = cv.getStructuringElement(cv.MORPH_RECT, (1,horizontalsize2))
erosion2 = cv.erode(th2_copy,horizontalStructure2,iterations = 1)
dilation2 = cv.dilate(erosion2,horizontalStructure2,iterations = 1)
#全横线图与全竖线图叠加，并提取交点
mask = dilation + dilation2
cv.imwrite('mask.jpg', mask)
joints = cv.bitwise_and(dilation, dilation2)
cv.imwrite('joints.jpg', joints)
#根据矩形大小筛选矩形框，并画在矫正后的表格上
#cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE
mask, contours, hierarchy = cv.findContours(mask,cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE)
length = len(contours)
print length
small_rects = []
big_rects = []
for i in range(length):
cnt = contours[i]
area = cv.contourArea(cnt)
if area < 10:
continue
approx = cv.approxPolyDP(cnt, 3, True)#3
x, y, w, h = cv.boundingRect(approx)
rect = (x, y, w, h)
#cv.rectangle(img, (x, y), (x+w, y+h), (0, 0, 255), 3)
roi = joints[y:y+h, x:x+w]
roi, joints_contours, joints_hierarchy = cv.findContours(roi,cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE)
#print len(joints_contours)
#if h < 80 and h > 20 and w > 10 and len(joints_contours)<=4:
if h < 80 and h > 20 and w > 10 and len(joints_contours)<=6:#important
cv.rectangle(image_copy, (x, y), (x+w, y+h), (255-h*3, h*3, 0), 3)
small_rects.append(rect)
cv.imwrite('table_out.jpg', image_copy)
##### 在不同类型的表格中根据信息所在的大致位置获取相应矩形框的坐标 #####
#矫正后的表格中信息的大致位置各在一定范围内，根据大致位置的坐标点筛选出该表中该信息对应的矩形框具体坐标
request_info_set = []#存储筛选出的所需矩形框坐标
#pure_for_message = shrinkTwoTimesTranslation.copy()
print "table_style:", table_style
if table_style == 1:
request_info = (770, 150, 1150-770, 300-150)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(770,150),(1150,300),(255,0,0),1)#登记表编号0
for j in range(0,len(small_rects)):
(x, y, w, h) = small_rects[j]
if x < 1060 and y < 1130 and x+w > 1060 and y+h > 1130 and something_missing == 1:#特殊情况下表格矩形的提取
x_rem = x
y_rem = y
w_rem = w
h_rem = h
for j in range(0,len(small_rects)):
(x, y, w, h) = small_rects[j]
if x < 700 and y < 370 and x+w > 700 and y+h > 370 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#性别1
if x < 880 and y < 370 and x+w > 880 and y+h > 370 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#民族2
if x < 685 and y < 412 and x+w > 685 and y+h > 412 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#体重3
if x < 880 and y < 412 and x+w > 880 and y+h > 412 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(0,0,255),1)#血型4
if x < 890 and y < 452 and x+w > 890 and y+h > 452 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#籍贯5
if x < 486 and y < 1090 and x+w > 486 and y+h > 1090 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中学校名称6
if x < 696 and y < 1090 and x+w > 696 and y+h > 1090 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中专业7
if x < 808 and y < 1090 and x+w > 808 and y+h > 1090 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中学位8
if x < 931 and y < 1090 and x+w > 931 and y+h > 1090 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中起止时间9
if x < 1060 and y < 1087 and x+w > 1060 and y+h > 1087 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中是否毕业10
#提取是、否选项旁的正方形框
if jiaozheng == 1 or (len(docu_num) == 9 and (docu_num[6] == '2' or docu_num[6] == '3')) or (len(docu_num) == 8 and ((docu_num[5] == '3' and (docu_num[6] == '3' or (docu_num[6] == '3' and (docu_num[7] == '5' or docu_num[7] == '6' or docu_num[7] == '7' or docu_num[7] == '8' or docu_num[7] == '9')) or docu_num[6] == '4' or docu_num[6] == '5' or docu_num[6] == '7' or docu_num[6] == '8' or docu_num[6] == '9')) or docu_num[5] == '4')):
gaozhong_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.18*w+x):int(0.28*w+x)]
gaozhong_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.59*w+x):int(0.69*w+x)]
cv.imwrite('gaozhong_roi_left.jpg', gaozhong_roi_left)
cv.imwrite('gaozhong_roi_right.jpg', gaozhong_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.18*w+x),int(y+h/3.5)),(int(0.28*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.59*w+x),int(y+h/3.5)),(int(0.69*w+x),int(y+2.2*h/3)),(255,0,0),1)
else:
gaozhong_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
gaozhong_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('gaozhong_roi_left.jpg', gaozhong_roi_left)
cv.imwrite('gaozhong_roi_right.jpg', gaozhong_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
if x < 486 and y < 1130 and x+w > 486 and y+h > 1130 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专学校名称11
if x < 696 and y < 1130 and x+w > 696 and y+h > 1130 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专专业12
if x < 808 and y < 1130 and x+w > 808 and y+h > 1130 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专学位13
if x < 931 and y < 1130 and x+w > 931 and y+h > 1130 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专起止时间14
if x < 1060 and y < 1127 and x+w > 1060 and y+h > 1127 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专是否毕业15
#提取是、否选项旁的正方形框
if jiaozheng == 1 or (len(docu_num) == 9 and (docu_num[6] == '2' or docu_num[6] == '3')) or (len(docu_num) == 8 and ((docu_num[5] == '3' and (docu_num[6] == '3' or (docu_num[6] == '3' and (docu_num[7] == '5' or docu_num[7] == '6' or docu_num[7] == '7' or docu_num[7] == '8' or docu_num[7] == '9')) or docu_num[6] == '4' or docu_num[6] == '5' or docu_num[6] == '7' or docu_num[6] == '8' or docu_num[6] == '9')) or docu_num[5] == '4')):
dazhuan_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.18*w+x):int(0.28*w+x)]
dazhuan_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.59*w+x):int(0.69*w+x)]
cv.imwrite('dazhuan_roi_left.jpg', dazhuan_roi_left)
cv.imwrite('dazhuan_roi_right.jpg', dazhuan_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.18*w+x),int(y+h/3.5)),(int(0.28*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.59*w+x),int(y+h/3.5)),(int(0.69*w+x),int(y+2.2*h/3)),(255,0,0),1)
else:
dazhuan_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
dazhuan_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('dazhuan_roi_left.jpg', dazhuan_roi_left)
cv.imwrite('dazhuan_roi_right.jpg', dazhuan_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
if x < 486 and y < 1170 and x+w > 486 and y+h > 1170 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科学校名称16
if x < 696 and y < 1170 and x+w > 696 and y+h > 1170 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科专业17
if x < 808 and y < 1170 and x+w > 808 and y+h > 1170 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科学位18
if x < 931 and y < 1170 and x+w > 931 and y+h > 1170 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科起止时间19
if x < 1060 and y < 1167 and x+w > 1060 and y+h > 1167 or something_missing == 1:
if something_missing == 1:#特殊情况下表格矩形的提取
x = x_rem
y = y_rem + 40
w = w_rem
h = h_rem
else:
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科是否毕业20
#提取是、否选项旁的正方形框
if jiaozheng == 1 or (len(docu_num) == 9 and (docu_num[6] == '2' or docu_num[6] == '3')) or (len(docu_num) == 8 and ((docu_num[5] == '3' and (docu_num[6] == '3' or (docu_num[6] == '3' and (docu_num[7] == '5' or docu_num[7] == '6' or docu_num[7] == '7' or docu_num[7] == '8' or docu_num[7] == '9')) or docu_num[6] == '4' or docu_num[6] == '5' or docu_num[6] == '7' or docu_num[6] == '8' or docu_num[6] == '9')) or docu_num[5] == '4')):
benke_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.18*w+x):int(0.28*w+x)]
benke_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.59*w+x):int(0.69*w+x)]
cv.imwrite('benke_roi_left.jpg', benke_roi_left)
cv.imwrite('benke_roi_right.jpg', benke_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.18*w+x),int(y+h/3.5)),(int(0.28*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.59*w+x),int(y+h/3.5)),(int(0.69*w+x),int(y+2.2*h/3)),(255,0,0),1)
else:
benke_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
benke_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('benke_roi_left.jpg', benke_roi_left)
cv.imwrite('benke_roi_right.jpg', benke_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
if x < 486 and y < 1210 and x+w > 486 and y+h > 1210 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生学校名称21
if x < 696 and y < 1210 and x+w > 696 and y+h > 1210 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生专业22
if x < 808 and y < 1210 and x+w > 808 and y+h > 1210 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生学位23
if x < 931 and y < 1210 and x+w > 931 and y+h > 1210 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生起止时间24
if x < 1060 and y < 1207 and x+w > 1060 and y+h > 1207 or something_missing == 1:
if something_missing == 1:#特殊情况下表格矩形的提取
x = x_rem
y = y_rem + 80
w = w_rem
h = h_rem
else:
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生是否毕业25
#提取是、否选项旁的正方形框
if jiaozheng == 1 or (len(docu_num) == 9 and (docu_num[6] == '2' or docu_num[6] == '3')) or (len(docu_num) == 8 and ((docu_num[5] == '3' and (docu_num[6] == '3' or (docu_num[6] == '3' and (docu_num[7] == '5' or docu_num[7] == '6' or docu_num[7] == '7' or docu_num[7] == '8' or docu_num[7] == '9')) or docu_num[6] == '4' or docu_num[6] == '5' or docu_num[6] == '7' or docu_num[6] == '8' or docu_num[6] == '9')) or docu_num[5] == '4')):
yanjiu_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.18*w+x):int(0.28*w+x)]
yanjiu_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.59*w+x):int(0.69*w+x)]
cv.imwrite('yanjiu_roi_left.jpg', yanjiu_roi_left)
cv.imwrite('yanjiu_roi_right.jpg', yanjiu_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.18*w+x),int(y+h/3.5)),(int(0.28*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.59*w+x),int(y+h/3.5)),(int(0.69*w+x),int(y+2.2*h/3)),(255,0,0),1)
else:
yanjiu_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
yanjiu_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('yanjiu_roi_left.jpg', yanjiu_roi_left)
cv.imwrite('yanjiu_roi_right.jpg', yanjiu_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
#特殊情况下表格矩形的提取
if need_stronger == 1:
cv.rectangle(shrinkTwoTimesTranslation,(673,385),(740,420),(0,0,255),1)#性别
cv.rectangle(shrinkTwoTimesTranslation,(845,385),(938,420),(0,0,255),1)#民族
cv.rectangle(shrinkTwoTimesTranslation,(673,425),(720,460),(0,0,255),1)#体重
cv.rectangle(shrinkTwoTimesTranslation,(845,425),(938,465),(0,0,255),1)#血型
cv.rectangle(shrinkTwoTimesTranslation,(845,470),(938,510),(0,0,255),1)#籍贯
cv.rectangle(shrinkTwoTimesTranslation,(973,1095),(1153,1135),(0,0,255),1)#高中是否毕业
x = 973
y = 1095
w = 1153 - 973
h = 1135 - 1095
gaozhong_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
gaozhong_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('gaozhong_roi_left.jpg', gaozhong_roi_left)
cv.imwrite('gaozhong_roi_right.jpg', gaozhong_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(973,1135),(1153,1175),(0,0,255),1)
x = 973
y = 1135
w = 1153 - 973
h = 1175 - 1135
dazhuan_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
dazhuan_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('dazhuan_roi_left.jpg', dazhuan_roi_left)
cv.imwrite('dazhuan_roi_right.jpg', dazhuan_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(973,1175),(1153,1215),(0,0,255),1)
x = 973
y = 1175
w = 1153 - 973
h = 1215 - 1175
benke_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
benke_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('benke_roi_left.jpg', benke_roi_left)
cv.imwrite('benke_roi_right.jpg', benke_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(973,1220),(1153,1260),(0,0,255),1)
x = 973
y = 1220
w = 1153 - 973
h = 1260 - 1220
yanjiu_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
yanjiu_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('yanjiu_roi_left.jpg', yanjiu_roi_left)
cv.imwrite('yanjiu_roi_right.jpg', yanjiu_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
#测试各信息的大致位置
#cv.circle(shrinkTwoTimesTranslation, (700, 335), 3,(255,0,0),3)
#cv.circle(shrinkTwoTimesTranslation, (75, 1581), 3,(255,0,0),3)
#cv.rectangle(shrinkTwoTimesTranslation,(663,360),(740,400),(255,0,0),1)#性别
#cv.rectangle(shrinkTwoTimesTranslation,(835,360),(928,400),(255,0,0),1)#民族
#cv.rectangle(shrinkTwoTimesTranslation,(663,402),(710,442),(255,0,0),1)#体重
#cv.rectangle(shrinkTwoTimesTranslation,(835,402),(928,442),(255,0,0),1)#血型
#cv.rectangle(shrinkTwoTimesTranslation,(835,443),(948,482),(255,0,0),1)#籍贯
#cv.rectangle(shrinkTwoTimesTranslation,(345,1075),(627,1115),(255,0,0),1)#高中学校名称
#cv.rectangle(shrinkTwoTimesTranslation,(628,1075),(765,1115),(255,0,0),1)#高中专业
#cv.rectangle(shrinkTwoTimesTranslation,(766,1075),(850,1115),(255,0,0),1)#高中学位
#cv.rectangle(shrinkTwoTimesTranslation,(851,1075),(1011,1115),(255,0,0),1)#高中起止时间
#cv.rectangle(shrinkTwoTimesTranslation,(971,1070),(1150,1110),(255,0,0),1)#高中是否毕业
if table_style == 2:
request_info = (770, 150, 1150-770, 300-150)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(770,150),(1150,300),(255,0,0),1)#登记表编号0
for j in range(0,len(small_rects)):
(x, y, w, h) = small_rects[j]
if x < 700 and y < 335 and x+w > 700 and y+h > 335 :#350-20
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#性别1
if x < 885 and y < 335 and x+w > 885 and y+h > 335 :#x+5
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#民族2
if x < 685 and y < 377 and x+w > 685 and y+h > 377 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#体重3
if x < 885 and y < 377 and x+w > 885 and y+h > 377 :#x+5
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(0,0,255),1)#血型4
if x < 890 and y < 417 and x+w > 890 and y+h > 417 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#籍贯5
if x < 486 and y < 1045 and x+w > 486 and y+h > 1045 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中学校名称6
if x < 696 and y < 1045 and x+w > 696 and y+h > 1045 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中专业7
if x < 808 and y < 1045 and x+w > 808 and y+h > 1045 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中学位8
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')):
y_xiuzheng = 1090
else:
y_xiuzheng = 1055
if x < 931 and y < y_xiuzheng and x+w > 931 and y+h > y_xiuzheng :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中起止时间9
if x < 1060 and y < y_xiuzheng and x+w > 1060 and y+h > y_xiuzheng :#y+10
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中是否毕业10
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')) or (docu_num[6] == '3' and docu_num[8] == '9'):
gaozhong_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.22*w+x):int(0.32*w+x)]
gaozhong_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.55*w+x):int(0.65*w+x)]
cv.imwrite('gaozhong_roi_left.jpg', gaozhong_roi_left)
cv.imwrite('gaozhong_roi_right.jpg', gaozhong_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.22*w+x),int(y+h/3.5)),(int(0.32*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.55*w+x),int(y+h/3.5)),(int(0.65*w+x),int(y+2.2*h/3)),(255,0,0),1)
else:
gaozhong_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
gaozhong_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('gaozhong_roi_left.jpg', gaozhong_roi_left)
cv.imwrite('gaozhong_roi_right.jpg', gaozhong_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
if x < 486 and y < 1085 and x+w > 486 and y+h > 1085 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专学校名称11
if x < 696 and y < 1085 and x+w > 696 and y+h > 1085 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专专业12
if x < 808 and y < 1085 and x+w > 808 and y+h > 1085 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专学位13
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')):
y_xiuzheng = 1140
else:
y_xiuzheng = 1095
if x < 931 and y < y_xiuzheng and x+w > 931 and y+h > y_xiuzheng :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专起止时间14
if x < 1060 and y < y_xiuzheng and x+w > 1060 and y+h > y_xiuzheng :#y+10
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专是否毕业15
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')) or (docu_num[6] == '3' and docu_num[8] == '9'):
dazhuan_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.22*w+x):int(0.32*w+x)]
dazhuan_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.55*w+x):int(0.65*w+x)]
cv.imwrite('dazhuan_roi_left.jpg', dazhuan_roi_left)
cv.imwrite('dazhuan_roi_right.jpg', dazhuan_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.22*w+x),int(y+h/3.5)),(int(0.32*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.55*w+x),int(y+h/3.5)),(int(0.65*w+x),int(y+2.2*h/3)),(255,0,0),1)
else:
dazhuan_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
dazhuan_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('dazhuan_roi_left.jpg', dazhuan_roi_left)
cv.imwrite('dazhuan_roi_right.jpg', dazhuan_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
if x < 486 and y < 1125 and x+w > 486 and y+h > 1125 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科学校名称16
if x < 696 and y < 1125 and x+w > 696 and y+h > 1125 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科专业17
if x < 808 and y < 1125 and x+w > 808 and y+h > 1125 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科学位18
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')):
y_xiuzheng = 1190
else:
y_xiuzheng = 1135
if x < 931 and y < y_xiuzheng and x+w > 931 and y+h > y_xiuzheng :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科起止时间19
if x < 1060 and y < y_xiuzheng and x+w > 1060 and y+h > y_xiuzheng :#y+10
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科是否毕业20
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')) or (docu_num[6] == '3' and docu_num[8] == '9'):
benke_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.22*w+x):int(0.32*w+x)]
benke_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.55*w+x):int(0.65*w+x)]
cv.imwrite('benke_roi_left.jpg', benke_roi_left)
cv.imwrite('benke_roi_right.jpg', benke_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.22*w+x),int(y+h/3.5)),(int(0.32*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.55*w+x),int(y+h/3.5)),(int(0.65*w+x),int(y+2.2*h/3)),(255,0,0),1)
else:
benke_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
benke_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('benke_roi_left.jpg', benke_roi_left)
cv.imwrite('benke_roi_right.jpg', benke_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
if x < 486 and y < 1165 and x+w > 486 and y+h > 1165 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生学校名称21
if x < 696 and y < 1165 and x+w > 696 and y+h > 1165 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生专业22
if x < 808 and y < 1165 and x+w > 808 and y+h > 1165 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生学位23
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')):
y_xiuzheng = 1230
else:
y_xiuzheng = 1175
if x < 931 and y < y_xiuzheng and x+w > 931 and y+h > y_xiuzheng :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生起止时间24
if x < 1060 and y < y_xiuzheng and x+w > 1060 and y+h > y_xiuzheng :#y+10
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生是否毕业25
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')) or (docu_num[6] == '3' and docu_num[8] == '9'):
yanjiu_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.22*w+x):int(0.32*w+x)]
yanjiu_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.55*w+x):int(0.65*w+x)]
cv.imwrite('yanjiu_roi_left.jpg', yanjiu_roi_left)
cv.imwrite('yanjiu_roi_right.jpg', yanjiu_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.22*w+x),int(y+h/3.5)),(int(0.32*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.55*w+x),int(y+h/3.5)),(int(0.65*w+x),int(y+2.2*h/3)),(255,0,0),1)
else:
yanjiu_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.28*w+x):int(0.38*w+x)]
yanjiu_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.2*h/3), int(0.51*w+x):int(0.61*w+x)]
cv.imwrite('yanjiu_roi_left.jpg', yanjiu_roi_left)
cv.imwrite('yanjiu_roi_right.jpg', yanjiu_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.28*w+x),int(y+h/3.5)),(int(0.38*w+x),int(y+2.2*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.51*w+x),int(y+h/3.5)),(int(0.61*w+x),int(y+2.2*h/3)),(255,0,0),1)
if table_style == 3:
request_info = (970, 50, 1200-970, 200-50)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(970,50),(1200,200),(255,0,0),1)#登记表编号0
for j in range(0,len(small_rects)):
(x, y, w, h) = small_rects[j]
if x < 724 and y < 316 and x+w > 724 and y+h > 316 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#性别1
if x < 960 and y < 316 and x+w > 960 and y+h > 316 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#民族2
if x < 724 and y < 363 and x+w > 724 and y+h > 363 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#体重3
if x < 960 and y < 363 and x+w > 960 and y+h > 363 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#血型4
if x < 960 and y < 410 and x+w > 960 and y+h > 410 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#籍贯5
if x < 981 and y < 1089 and x+w > 981 and y+h > 1089 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中起止时间9
if x < 1105 and y < 1089 and x+w > 1105 and y+h > 1089 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中是否毕业10
if docu_num[5] == '3' and docu_num[6] == '8':
gaozhong_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.19*w+x):int(0.29*w+x)]
gaozhong_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.59*w+x):int(0.69*w+x)]
cv.imwrite('gaozhong_roi_left.jpg', gaozhong_roi_left)
cv.imwrite('gaozhong_roi_right.jpg', gaozhong_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.19*w+x),int(y+h/3.5)),(int(0.29*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.59*w+x),int(y+h/3.5)),(int(0.69*w+x),int(y+2.1*h/3)),(255,0,0),1)
else:
gaozhong_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.23*w+x):int(0.33*w+x)]
gaozhong_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.56*w+x):int(0.66*w+x)]
cv.imwrite('gaozhong_roi_left.jpg', gaozhong_roi_left)
cv.imwrite('gaozhong_roi_right.jpg', gaozhong_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.23*w+x),int(y+h/3.5)),(int(0.33*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.56*w+x),int(y+h/3.5)),(int(0.66*w+x),int(y+2.1*h/3)),(255,0,0),1)
if x < 981 and y < 1137 and x+w > 981 and y+h > 1137 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专起止时间14
if x < 1105 and y < 1137 and x+w > 1105 and y+h > 1137 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专是否毕业15
if docu_num[5] == '3' and docu_num[6] == '8':
dazhuan_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.19*w+x):int(0.29*w+x)]
dazhuan_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.59*w+x):int(0.69*w+x)]
cv.imwrite('dazhuan_roi_left.jpg', dazhuan_roi_left)
cv.imwrite('dazhuan_roi_right.jpg', dazhuan_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.19*w+x),int(y+h/3.5)),(int(0.29*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.59*w+x),int(y+h/3.5)),(int(0.69*w+x),int(y+2.1*h/3)),(255,0,0),1)
else:
dazhuan_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.23*w+x):int(0.33*w+x)]
dazhuan_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.56*w+x):int(0.66*w+x)]
cv.imwrite('dazhuan_roi_left.jpg', dazhuan_roi_left)
cv.imwrite('dazhuan_roi_right.jpg', dazhuan_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.23*w+x),int(y+h/3.5)),(int(0.33*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.56*w+x),int(y+h/3.5)),(int(0.66*w+x),int(y+2.1*h/3)),(255,0,0),1)
if x < 981 and y < 1185 and x+w > 981 and y+h > 1185 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科起止时间19
if x < 1105 and y < 1185 and x+w > 1105 and y+h > 1185 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科是否毕业20
if docu_num[5] == '3' and docu_num[6] == '8':
benke_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.19*w+x):int(0.29*w+x)]
benke_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.59*w+x):int(0.69*w+x)]
cv.imwrite('benke_roi_left.jpg', benke_roi_left)
cv.imwrite('benke_roi_right.jpg', benke_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.19*w+x),int(y+h/3.5)),(int(0.29*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.59*w+x),int(y+h/3.5)),(int(0.69*w+x),int(y+2.1*h/3)),(255,0,0),1)
else:
benke_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.23*w+x):int(0.33*w+x)]
benke_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.56*w+x):int(0.66*w+x)]
cv.imwrite('benke_roi_left.jpg', benke_roi_left)
cv.imwrite('benke_roi_right.jpg', benke_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.23*w+x),int(y+h/3.5)),(int(0.33*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.56*w+x),int(y+h/3.5)),(int(0.66*w+x),int(y+2.1*h/3)),(255,0,0),1)
if x < 981 and y < 1233 and x+w > 981 and y+h > 1233 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生起止时间24
if x < 1105 and y < 1233 and x+w > 1105 and y+h > 1233 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生是否毕业25
if docu_num[5] == '3' and docu_num[6] == '8':
yanjiu_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.19*w+x):int(0.29*w+x)]
yanjiu_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.59*w+x):int(0.69*w+x)]
cv.imwrite('yanjiu_roi_left.jpg', yanjiu_roi_left)
cv.imwrite('yanjiu_roi_right.jpg', yanjiu_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.19*w+x),int(y+h/3.5)),(int(0.29*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.59*w+x),int(y+h/3.5)),(int(0.69*w+x),int(y+2.1*h/3)),(255,0,0),1)
else:
yanjiu_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.23*w+x):int(0.33*w+x)]
yanjiu_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.56*w+x):int(0.66*w+x)]
cv.imwrite('yanjiu_roi_left.jpg', yanjiu_roi_left)
cv.imwrite('yanjiu_roi_right.jpg', yanjiu_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.23*w+x),int(y+h/3.5)),(int(0.33*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.56*w+x),int(y+h/3.5)),(int(0.66*w+x),int(y+2.1*h/3)),(255,0,0),1)
if table_style == 4:
request_info = (970, 50, 1200-970, 200-50)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(970,50),(1200,200),(255,0,0),1)#登记表编号0
for j in range(0,len(small_rects)):
(x, y, w, h) = small_rects[j]
if x < 724 and y < 292 and x+w > 724 and y+h > 292 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#性别1
if x < 960 and y < 292 and x+w > 960 and y+h > 292 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#民族2
if x < 724 and y < 339 and x+w > 724 and y+h > 339 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#体重3
if x < 960 and y < 339 and x+w > 960 and y+h > 339 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#血型4
if x < 960 and y < 386 and x+w > 960 and y+h > 386 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#籍贯5
if x < 981 and y < 1094 and x+w > 981 and y+h > 1094 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中起止时间9
if x < 1105 and y < 1094 and x+w > 1105 and y+h > 1094 :#y-15
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#高中是否毕业10
gaozhong_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.23*w+x):int(0.33*w+x)]
gaozhong_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.56*w+x):int(0.66*w+x)]
cv.imwrite('gaozhong_roi_left.jpg', gaozhong_roi_left)
cv.imwrite('gaozhong_roi_right.jpg', gaozhong_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.23*w+x),int(y+h/3.5)),(int(0.33*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.56*w+x),int(y+h/3.5)),(int(0.66*w+x),int(y+2.1*h/3)),(255,0,0),1)
if x < 981 and y < 1142 and x+w > 981 and y+h > 1142 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专起止时间14
if x < 1105 and y < 1142 and x+w > 1105 and y+h > 1142 :#y-15
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#大专是否毕业15
dazhuan_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.23*w+x):int(0.33*w+x)]
dazhuan_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.56*w+x):int(0.66*w+x)]
cv.imwrite('dazhuan_roi_left.jpg', dazhuan_roi_left)
cv.imwrite('dazhuan_roi_right.jpg', dazhuan_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.23*w+x),int(y+h/3.5)),(int(0.33*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.56*w+x),int(y+h/3.5)),(int(0.66*w+x),int(y+2.1*h/3)),(255,0,0),1)
if x < 981 and y < 1190 and x+w > 981 and y+h > 1190 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科起止时间19
if x < 1105 and y < 1190 and x+w > 1105 and y+h > 1190 :#y-15
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#本科是否毕业20
benke_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.23*w+x):int(0.33*w+x)]
benke_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.56*w+x):int(0.66*w+x)]
cv.imwrite('benke_roi_left.jpg', benke_roi_left)
cv.imwrite('benke_roi_right.jpg', benke_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.23*w+x),int(y+h/3.5)),(int(0.33*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.56*w+x),int(y+h/3.5)),(int(0.66*w+x),int(y+2.1*h/3)),(255,0,0),1)
if x < 981 and y < 1238 and x+w > 981 and y+h > 1238 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生起止时间24
if x < 1105 and y < 1238 and x+w > 1105 and y+h > 1238 :#y-15
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#研究生是否毕业25
yanjiu_roi_left = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.23*w+x):int(0.33*w+x)]
yanjiu_roi_right = shrinkTwoTimesTranslation[int(y+h/3.5):int(y+2.1*h/3), int(0.56*w+x):int(0.66*w+x)]
cv.imwrite('yanjiu_roi_left.jpg', yanjiu_roi_left)
cv.imwrite('yanjiu_roi_right.jpg', yanjiu_roi_right)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.23*w+x),int(y+h/3.5)),(int(0.33*w+x),int(y+2.1*h/3)),(255,0,0),1)
cv.rectangle(shrinkTwoTimesTranslation,(int(0.56*w+x),int(y+h/3.5)),(int(0.66*w+x),int(y+2.1*h/3)),(255,0,0),1)
#cv.rectangle(shrinkTwoTimesTranslation,(663,269),(786,315),(255,0,0),1)#性别
#cv.rectangle(shrinkTwoTimesTranslation,(900,269),(1020,315),(255,0,0),1)#民族
#cv.rectangle(shrinkTwoTimesTranslation,(663,316),(786,362),(255,0,0),1)#体重
#cv.rectangle(shrinkTwoTimesTranslation,(900,316),(1020,362),(255,0,0),1)#血型
#cv.rectangle(shrinkTwoTimesTranslation,(900,363),(1020,409),(255,0,0),1)#籍贯
#cv.rectangle(shrinkTwoTimesTranslation,(1016,1065),(1195,1113),(255,0,0),1)#高中是否毕业
#登记表编号,性别,民族,体重,血型,籍贯,高中学校名称,高中专业,高中学位,高中起止时间,高中是否毕业,大专学校名称,大专专业,大专学位,大专起止时间,大专是否毕业,本科学校名称,本科专业,本科学位,本科起止时间,本科是否毕业,研究生学校名称,研究生专业,研究生学位,研究生起止时间,研究生是否毕业,学校名称（其它）,专业（其它）,学位（其它）,起止时间（其它）,是否毕业（其它）
cv.imwrite('image_correct.jpg', shrinkTwoTimesTranslation)#关键信息框重点标注的矫正图像
########### 去除噪点Filter ###########
th2 = cv.medianBlur(th2,5)#5
horizontalStructure = cv.getStructuringElement(cv.MORPH_RECT, (2, 2))#2,2
th2 = cv.erode(th2,horizontalStructure,iterations = 1)
#horizontalStructure = cv.getStructuringElement(cv.MORPH_RECT, (1, 3))#2,2
#th2 = cv.dilate(th2,horizontalStructure,iterations = 1)
########### 登记表编号0 ###########
#(x, y, w, h) = request_info_set[0]
#num_num = 8
#num_result = num_out.num_o(th2, x, y, w, h, num_num)
#print "登记表编号: ",num_result
########### 识别 性别1 ###########
if table_style == 1:
x_com = 700#信息所在的大致位置
y_com = 370
if table_style == 2:
x_com = 700
y_com = 335
if table_style == 3:
x_com = 724
y_com = 316
if table_style == 4:
x_com = 724
y_com = 292
if need_stronger == 1:#特殊情况处理
x = 673
y = 385
w = 740 - 673
h = 420 - 385
get_all = 1
cha_num = 3
is_ABO = 0
xingbie_result = ''
return_change, xingbie_result_set = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
print xingbie_result_set
if xingbie_result_set[0] == 2206 or xingbie_result_set[1] == 2206 or xingbie_result_set[2] == 2206:
xingbie_result = '男'
if xingbie_result_set[0] == 775 or xingbie_result_set[1] == 775 or xingbie_result_set[2] == 775:
xingbie_result = '女'
if xingbie_result == '':
xingbie_result = '男'
print "性别: ",xingbie_result
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :#提取信息所在的矩形框
get_all = 1#对手写汉字模型的三个识别结果都进行分析
cha_num = 3#字段预计最长长度
is_ABO = 0#不是血型数据
xingbie_result = ''#初始化
#调用chinese_out.py中的chinese_o()函数进行手写汉字识别
return_change, xingbie_result_set = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
print xingbie_result_set
#如果三个识别结果中有一个'男'，则确定为男性
if xingbie_result_set[0] == 2206 or xingbie_result_set[1] == 2206 or xingbie_result_set[2] == 2206:
xingbie_result = '男'
#如果三个识别结果中有一个'女'，则确定为女性
if xingbie_result_set[0] == 775 or xingbie_result_set[1] == 775 or xingbie_result_set[2] == 775:
xingbie_result = '女'
#特殊情况处理
if xingbie_result == '':
xingbie_result = '男'
print "性别: ",xingbie_result
########### 识别 民族2 ###########
if table_style == 1:
x_com = 880
y_com = 370
if table_style == 2:
x_com = 885
y_com = 335
if table_style == 3:
x_com = 960
y_com = 316
if table_style == 4:
x_com = 960
y_com = 292
minzu_result = '汉'
if need_stronger == 1:
x = 845
y = 385
w = 938 - 845
h = 420 - 385
get_all = 2
cha_num = 1
is_ABO = 0
return_change, minzu_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
if return_change != 0:
if minzu_result[0] == 1809 or minzu_result[1] == 1809 or minzu_result[2] == 1809:
minzu_result = '汉'
else:
minzu_result = return_change
print "民族: ",minzu_result
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :
get_all = 2#对手写汉字模型的三个识别结果都进行分析，且标志当前为民族数据提取
cha_num = 1#字段预计最长长度
is_ABO = 0#不是血型数据
#调用chinese_out.py中的chinese_o()函数进行手写汉字识别
return_change, minzu_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
if return_change != 0:#特殊情况处理
#如果三个识别结果中有一个'汉'，则确定为汉，否则输出单字
if minzu_result[0] == 1809 or minzu_result[1] == 1809 or minzu_result[2] == 1809:
minzu_result = '汉'
else:
minzu_result = return_change
print "民族: ",minzu_result
print "民族: ",minzu_result
########### 识别 体重3 ###########
if table_style == 1:
x_com = 685
y_com = 412
if table_style == 2:
x_com = 685
y_com = 377
if table_style == 3:
x_com = 724
y_com = 363
if table_style == 4:
x_com = 724
y_com = 339
if need_stronger == 1:
x = 673
y = 425
w = 720 - 673
h = 460 - 425
num_num = 3
weight_result = num_out.num_o(th2_copy, x, y, int(w*0.6), h, num_num)
if weight_result == '':
weight_result = '无'
print "体重: ",weight_result
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :
num_num = 3#字段预计最长长度
#调用num_out.py中的num_o()函数进行手写数字识别，体重框只取前60%部分（避免kg干扰）
weight_result = num_out.num_o(th2_copy, x, y, int(w*0.6), h, num_num)#0.6
#特殊情况处理
if weight_result == '':
weight_result = '无'
if len(weight_result) > 3:
weight_result = weight_result[0:2]
#if weight_result[0] != '1':
#weight_result = weight_result[0:1]
print "体重: ",weight_result
########### 识别 血型4 ###########
if table_style == 1:
x_com = 880
y_com = 412
if table_style == 2:
x_com = 885
y_com = 377
if table_style == 3:
x_com = 960
y_com = 363
if table_style == 4:
x_com = 960
y_com = 339
xuexing_result = 'O'
if need_stronger == 1:
x = 845
y = 425
w = 938 - 845
h = 465 - 425
get_all = 0
cha_num = 2
is_ABO = 1
return_change, xuexing_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
print xuexing_result, len(xuexing_result)
if xuexing_result == '':
xuexing_result = '无'
if len(xuexing_result) >= 3 and (xuexing_result[2] == 'A' or xuexing_result[2] == 'B' or xuexing_result[2] == 'O'):
if return_change != 2:
xuexing_result = xuexing_result[0]
if len(xuexing_result) == 2 and return_change != 2:
xuexing_result = 'AB'
print "血型: ",xuexing_result
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :
get_all = 0#只分析模型的最佳识别结果
cha_num = 2#字段预计最长长度
is_ABO = 1#是血型数据
#调用chinese_out.py中的chinese_o()函数进行手写血型识别
return_change, xuexing_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
print xuexing_result, len(xuexing_result)
#特殊情况处理
if xuexing_result == '':
xuexing_result = '无'
if len(xuexing_result) >= 3 and (xuexing_result[2] == 'A' or xuexing_result[2] == 'B' or xuexing_result[2] == 'O'):
if return_change != 2:
xuexing_result = xuexing_result[0]
if len(xuexing_result) == 2 and return_change != 2:
xuexing_result = 'AB'
print "血型: ",xuexing_result
print "血型: ",xuexing_result
########### 识别 籍贯5 ###########
jiguan_result = '无'
if table_style == 1:
x_com = 890
y_com = 452
if table_style == 2:
x_com = 890
y_com = 417
if table_style == 3:
x_com = 960
y_com = 410
if table_style == 4:
x_com = 960
y_com = 386
jiguan_result = ''
try:
if need_stronger == 1:
x = 845
y = 465
w = 938 - 845
h = 465 - 425
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 3
is_ABO = 0
return_change, jiguan_result = chinese_out_jiguan.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
jiguan_result = '无'
print "籍贯: ",jiguan_result
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:#排除空框
get_all = 0
cha_num = 6#字段预计最长长度
is_ABO = 0
#调用chinese_out_jiguan.py中的chinese_o()函数进行手写汉字识别
return_change, jiguan_result = chinese_out_jiguan.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
jiguan_result = '无'
print "籍贯: ",jiguan_result
except:
jiguan_result = '无'
print "籍贯: ",jiguan_result
########### 识别 高中学校名称6 ###########
gaozhong_name_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 486 and y < 1090 and x+w > 486 and y+h > 1090 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 12
is_ABO = 0
gaozhong_name_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
gaozhong_name_result = '无'
print "高中学校名称: ",gaozhong_name_result
'''
########### 识别 高中专业7 ###########
gaozhong_master_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 696 and y < 1090 and x+w > 696 and y+h > 1090 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 6
is_ABO = 0
gaozhong_master_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
gaozhong_master_result = '无'
print "高中专业: ",gaozhong_master_result
'''
########### 识别 高中学位8 ###########
gaozhong_degree_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 808 and y < 1090 and x+w > 808 and y+h > 1090 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 3
is_ABO = 0
gaozhong_degree_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
gaozhong_degree_result = '无'
print "高中学位: ",gaozhong_degree_result
'''
########### 识别 高中起止时间9 ###########
gaozhong_time_result = '无'
if table_style == 1:
x_com = 931
y_com = 1090
if table_style == 2:#特殊情况处理
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')):
y_com = 1090
else:
y_com = 1055
x_com = 931
if table_style == 3:
x_com = 981
y_com = 1089
if table_style == 4:
x_com = 981
y_com = 1094
try:
if need_stronger == 1:
gaozhong_time_result = '无'
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:#排除空框
num_num = 10#字段预计最长长度
#调用time_out.py中的num_o()函数进行手写数字识别
gaozhong_time_result = time_out.num_o(th2_copy, x, y, w, h, num_num)
else:
gaozhong_time_result = '无'
print "高中起止时间: ",gaozhong_time_result
except:
gaozhong_time_result = '无'
print "高中起止时间: ",gaozhong_time_result
########### 识别 高中是否毕业10 ###########
left_mean = cv.mean(gaozhong_roi_left)#求提取出的正方形区域均值
right_mean = cv.mean(gaozhong_roi_right)
print left_mean[0],right_mean[0]
mean_diff = 14#两者间的差值超过该值则判断为是、否，否则判断为两个空框
if left_mean[0] < right_mean[0] - mean_diff:
gaozhong_biye_result = '是'
else:
if right_mean[0] < left_mean[0] - mean_diff:
gaozhong_biye_result = '否'
else:
gaozhong_biye_result = '无'
print "高中是否毕业: ",gaozhong_biye_result
########### 识别 大专学校名称11 ###########
dazhuan_name_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 486 and y < 1130 and x+w > 486 and y+h > 1130 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 12
is_ABO = 0
dazhuan_name_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
dazhuan_name_result = '无'
print "大专学校名称: ",dazhuan_name_result
'''
########### 识别 大专专业12 ###########
dazhuan_master_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 696 and y < 1130 and x+w > 696 and y+h > 1130 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 6
is_ABO = 0
dazhuan_master_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
dazhuan_master_result = '无'
print "大专专业: ",dazhuan_master_result
'''
########### 识别 大专学位13 ###########
dazhuan_degree_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 808 and y < 1130 and x+w > 808 and y+h > 1130 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 3
is_ABO = 0
dazhuan_degree_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
dazhuan_degree_result = '无'
print "大专学位: ",dazhuan_degree_result
'''
########### 识别 大专起止时间14 ###########
dazhuan_time_result = '无'
if table_style == 1:
x_com = 931
y_com = 1130
if table_style == 2:
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')):
y_com = 1140
else:
y_com = 1095
x_com = 931
if table_style == 3:
x_com = 981
y_com = 1137
if table_style == 4:
x_com = 981
y_com = 1142
try:
if need_stronger == 1:
dazhuan_time_result = '无'
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
num_num = 10
#调用time_out.py中的num_o()函数进行手写数字识别
dazhuan_time_result = time_out.num_o(th2_copy, x, y, w, h, num_num)
else:
dazhuan_time_result = '无'
print "大专起止时间: ",dazhuan_time_result
except:
dazhuan_time_result = '无'
print "大专起止时间: ",dazhuan_time_result
########### 识别 大专是否毕业15 ###########
left_mean = cv.mean(dazhuan_roi_left)
right_mean = cv.mean(dazhuan_roi_right)
print left_mean[0],right_mean[0]
mean_diff = 14#两者间的差值超过该值则判断为是、否，否则判断为两个空框
if left_mean[0] < right_mean[0] - mean_diff:
dazhuan_biye_result = '是'
else:
if right_mean[0] < left_mean[0] - mean_diff:
dazhuan_biye_result = '否'
else:
dazhuan_biye_result = '无'
print "大专是否毕业: ",dazhuan_biye_result
########### 识别 本科学校名称16 ###########
benke_name_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 486 and y < 1170 and x+w > 486 and y+h > 1170 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 12
is_ABO = 0
benke_name_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
benke_name_result = '无'
print "本科学校名称: ",benke_name_result
'''
########### 识别 本科专业17 ###########
benke_master_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 696 and y < 1170 and x+w > 696 and y+h > 1170 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 6
is_ABO = 0
benke_master_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
benke_master_result = '无'
print "本科专业: ",benke_master_result
'''
########### 识别 本科学位18 ###########
benke_degree_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 808 and y < 1170 and x+w > 808 and y+h > 1170 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 3
is_ABO = 0
benke_degree_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
benke_degree_result = '无'
print "本科学位: ",benke_degree_result
'''
########### 识别 本科起止时间19 ###########
benke_time_result = '无'
if table_style == 1:
x_com = 931
y_com = 1170
if table_style == 2:
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')):
y_com = 1190
else:
y_com = 1135
x_com = 931
if table_style == 3:
x_com = 981
y_com = 1185
if table_style == 4:
x_com = 981
y_com = 1190
try:
if need_stronger == 1:
benke_time_result = '无'
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
num_num = 10
#调用time_out.py中的num_o()函数进行手写数字识别
benke_time_result = time_out.num_o(th2_copy, x, y, w, h, num_num)
else:
benke_time_result = '无'
print "本科起止时间: ",benke_time_result
except:
benke_time_result = '无'
print "本科起止时间: ",benke_time_result
########### 识别 本科是否毕业20 ###########
left_mean = cv.mean(benke_roi_left)
right_mean = cv.mean(benke_roi_right)
print left_mean[0],right_mean[0]
mean_diff = 14#两者间的差值超过该值则判断为是、否，否则判断为两个空框
if left_mean[0] < right_mean[0] - mean_diff:
benke_biye_result = '是'
else:
if right_mean[0] < left_mean[0] - mean_diff:
benke_biye_result = '否'
else:
benke_biye_result = '无'
print "本科是否毕业: ",benke_biye_result
########### 识别 研究生学校名称21 ###########
yanjiu_name_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 486 and y < 1210 and x+w > 486 and y+h > 1210 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 12
is_ABO = 0
yanjiu_name_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
yanjiu_name_result = '无'
print "研究生学校名称: ",yanjiu_name_result
'''
########### 识别 研究生专业22 ###########
yanjiu_master_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 696 and y < 1210 and x+w > 696 and y+h > 1210 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 6
is_ABO = 0
yanjiu_master_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
yanjiu_master_result = '无'
print "研究生专业: ",yanjiu_master_result
'''
########### 识别 研究生学位23 ###########
yanjiu_degree_result = '无'#功能已实现，因效果一般，暂时注释
'''
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < 808 and y < 1210 and x+w > 808 and y+h > 1210 :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
get_all = 0
cha_num = 3
is_ABO = 0
yanjiu_degree_result = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
else:
yanjiu_degree_result = '无'
print "研究生学位: ",yanjiu_degree_result
'''
########### 识别 研究生起止时间24 ###########
yanjiu_time_result = '无'
if table_style == 1:
x_com = 931
y_com = 1210
if table_style == 2:
if len(docu_num) == 9 and (docu_num[6] == '2' and (docu_num[7] == '6' or docu_num[7] == '8' or docu_num[7] == '9')):
y_com = 1230
else:
y_com = 1175
x_com = 931
if table_style == 3:
x_com = 981
y_com = 1233
if table_style == 4:
x_com = 981
y_com = 1238
try:
if need_stronger == 1:
yanjiu_time_result = '无'
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :
pic_mean = cv.mean(th2_copy[y:y+h, x:x+w])
print "pic_mean[0]:", pic_mean[0]
if pic_mean[0] > 30:
num_num = 10
#调用time_out.py中的num_o()函数进行手写数字识别
yanjiu_time_result = time_out.num_o(th2_copy, x, y, w, h, num_num)
else:
yanjiu_time_result = '无'
print "研究生起止时间: ",yanjiu_time_result
except:
yanjiu_time_result = '无'
print "研究生起止时间: ",yanjiu_time_result
########### 识别 研究生是否毕业25 ###########
left_mean = cv.mean(yanjiu_roi_left)
right_mean = cv.mean(yanjiu_roi_right)
print left_mean[0],right_mean[0]
mean_diff = 14#两者间的差值超过该值则判断为是、否，否则判断为两个空框
if left_mean[0] < right_mean[0] - mean_diff:
yanjiu_biye_result = '是'
else:
if right_mean[0] < left_mean[0] - mean_diff:
yanjiu_biye_result = '否'
else:
yanjiu_biye_result = '无'
print "研究生是否毕业: ",yanjiu_biye_result
########### 识别 其它学校名称26 ###########
qita_name_result = '无'
########### 识别 其它专业27 ###########
qita_master_result = '无'
########### 识别 其它学位28 ###########
qita_degree_result = '无'
########### 识别 其它起止时间29 ###########
qita_time_result = '无'
########### 识别 其它是否毕业30 ###########
qita_biye_result = '无'
##### 向CSV文件写入数据 #####
information = [docu_num, xingbie_result, minzu_result, weight_result, xuexing_result, jiguan_result, gaozhong_name_result, gaozhong_master_result, gaozhong_degree_result, gaozhong_time_result, gaozhong_biye_result, dazhuan_name_result, dazhuan_master_result, dazhuan_degree_result, dazhuan_time_result, dazhuan_biye_result, benke_name_result, benke_master_result, benke_degree_result, benke_time_result, benke_biye_result, yanjiu_name_result, yanjiu_master_result, yanjiu_degree_result, yanjiu_time_result, yanjiu_biye_result, qita_name_result, qita_master_result, qita_degree_result, qita_time_result, qita_biye_result]
csvwrite.writerow(information)
够长吧哈哈，我们来分解一下上述代码中校正并提取表格中各个框的部分：
1、import包及汉字编码
#!/usr/bin/env python
# encoding: utf-8
import os
import cv2 as cv
import numpy as np
import math如果python文件中含有汉字（包括注释），就一定要有#encoding:utf-8这一行，否则会报xe5的错。不要看有个注释符，其实这句话还是起作用的。import cv2 as cv即添加OpenCV的Python包，使用cv.()的方式即可调用OpenCV的函数，类似C++中的cv::()的形式。
2、读取表格图片文件
path = 'test_data'
for (path,dirs,files) in os.walk(path):
for filename in files:
#docu_num = '20110158' #测试单张登记表
##### 按文件名读取登记表 #####
(docu_num,extension) = os.path.splitext(filename)
print "filename:", filename这段即将test_data文件中的所有文件进行遍历读取，也可注释最下方的两行，并去除docu_num一行的注释，即可对一张图片文件进行反复的读取、处理。
3、统计图中长横线的斜率来判断整体需要旋转矫正的角度
image = cv.imread('test_data/' + docu_num + '.jpg')
rows, cols, channels = image.shape
print rows, cols
image_copy = image.copy()
##### 旋转校正Rotation #####
#统计图中长横线的斜率来判断整体需要旋转矫正的角度
gray = cv.cvtColor(image, cv.COLOR_RGB2GRAY)
if table_style == 1 or table_style == 3 or table_style == 2:
edges = cv.Canny(gray, 50, 150, apertureSize=3) # 50,150,3
cv.imwrite('edges_whole.jpg', edges)
lines = cv.HoughLinesP(edges, 1, np.pi / 180, 500, 0, minLineLength=50, maxLineGap=50)#650,50,20
if table_style == 4:
edges_gray = cv.Canny(gray, 50, 150, apertureSize=3) # 50,150,3
edges = edges_gray[400:1000, 0:1000]
cv.imwrite('edges_whole.jpg', edges)
lines = cv.HoughLinesP(edges, 1, np.pi / 180, 200, 0, minLineLength=50, maxLineGap=35)#650,50,20
pi = 3.1415
theta_total = 0
theta_count = 0
for line in lines:
x1, y1, x2, y2 = line[0]
if table_style == 4:
y1 = y1 + 400
y2 = y2 + 400
rho = np.sqrt((x1 - x2)**2 + (y1 - y2)**2)
theta = math.atan(float(y2 - y1)/float(x2 - x1 + 0.001))
print(rho, theta, x1, y1, x2, y2)
if theta < pi/4 and theta > -pi/4:
theta_total = theta_total + theta
theta_count+=1
cv.line(image_copy, (x1, y1), (x2, y2), (0, 0, 255), 2)
#cv.line(edges, (x1, y1), (x2, y2), (0, 0, 0), 2)
theta_average = theta_total/theta_count
print theta_average, theta_average*180/pi
cv.imwrite('line_detect4rotation.jpg', image_copy)
#cv.imwrite('line_detect4rotation.jpg', ~edges)
#affineShrinkTranslationRotation = cv.getRotationMatrix2D((cols/2, rows/2), theta_average*180/pi, 1)
affineShrinkTranslationRotation = cv.getRotationMatrix2D((0, rows), theta_average*180/pi, 1)
ShrinkTranslationRotation = cv.warpAffine(image, affineShrinkTranslationRotation, (cols, rows))
image_copy = cv.warpAffine(image_copy, affineShrinkTranslationRotation, (cols, rows))
cv.imwrite('image_Rotation.jpg',ShrinkTranslationRotation)这段做的是旋转校正操作，先把表格图片转换为灰度图，再用Canny算子提取边缘（灰度+Canny是提取边缘的标准操作）得到如下这张图片edges_whole.jpg：
可见，这张表格是倾斜的，需要对表格进行旋转。旋转需要参照物吧？笔者选择用霍夫变换HoughLinesP()对表格中的长直线进行识别并提取。然后再对这些长直线中斜率小于pi/4且大于-pi/4的直线进行筛选，统计它们的斜率平均值。这样除非这张表格的倾斜度超过45度，或者表格中含有人为划的长横线，这个斜率平均值都可以作为校正旋转的角度了。长直线筛选后的图片见下图line_detect4rotation.jpg：
之后便是用getRotationMatrix2D()、warpAffine()函数进行旋转变换，此处笔者拷贝了一份不进行绘图操作的图片（不然都涂花了，干嘛为难自己）。旋转后的表格图片如下image_Rotation.jpg：
4、通过对表格左下角直角进行识别，将其顶点统一平移矫正至(78,1581)
##### 平移校正Move #####
#通过对表格左下角直角进行识别，将其顶点统一平移矫正至(78,1581)
#print "rows: ",rows
roi = image_copy[1450:rows, 0:150]#180
gray = cv.cvtColor(roi, cv.COLOR_RGB2GRAY)
edges = cv.Canny(gray, 50, 150, apertureSize=3) # 50,150,3
roi_mean_set = cv.mean(~edges[0:int((rows-1450)/2), 85:150])#通过区域灰度值特征排除文字对直线识别的干扰
roi_mean = roi_mean_set[0]
#cv.imwrite('edges_sample.jpg', edges)
cv.imwrite('edges_sample.jpg', ~edges[0:int((rows-1450)/2), 75:150])
lines = cv.HoughLinesP(edges, 1.0, np.pi / 180, 35, 0, minLineLength=10,maxLineGap=20)#50,10,20
lines_message_set = []
for line in lines:
x1, y1, x2, y2 = line[0]
y1 = y1 + 1450
y2 = y2 + 1450
rho = np.sqrt((x1 - x2)**2 + (y1 - y2)**2)
xielv = (y2 - y1)/(x2 - x1 + 0.001)
theta = math.atan(float(y2 - y1)/float(x2 - x1 + 0.001))
print(rho, theta, x1, y1, x2, y2, xielv)
lines_message = (rho, theta, x1, y1, x2, y2, xielv)
#cv.line(image, (x1, y1), (x2, y2), (0, 0, 255), 2)
lines_message_set.append(lines_message)
#print len(lines_message_set)
#求点到直线的距离
def point2line_distance(x1,y1,x2,y2,pointPx,pointPy):
A = y1 - y2
B = x2 - x1
C = x1*y2 - y1*x2
distance = abs(A*pointPx + B*pointPy + C)/((A*A + B*B)**0.5)
return distance
lines_cluster_set = []
repeat_num_set = []
for j in range(0,len(lines_message_set)):
for i in range(0,len(lines_message_set)):
if not(j in repeat_num_set):
lines_cluster_set.append(lines_message_set[j])
repeat_num_set.append(j)
print point2line_distance(lines_message_set[i][2], lines_message_set[i][3], lines_message_set[i][4], lines_message_set[i][5], (lines_message_set[j][2]+lines_message_set[j][4])/2, (lines_message_set[j][3]+lines_message_set[j][5])/2)
if i!=j and abs(lines_message_set[j][6] - lines_message_set[i][6]) < 0.1 and point2line_distance(lines_message_set[i][2], lines_message_set[i][3], lines_message_set[i][4], lines_message_set[i][5], (lines_message_set[j][2]+lines_message_set[j][4])/2, (lines_message_set[j][3]+lines_message_set[j][5])/2) <= 10:
repeat_num_set.append(i)
print lines_cluster_set
#对直角的横线、竖线进行分析，缺省时根据表格类型进行矫正
Point_heng = []
Point_shu = []
MiddlePoint_heng = (112,1450+(rows-1450)/2)
MiddlePoint_shu = (75,1450+(rows-1450)/4)
distance2point = rows-1450
distance2point2 = rows-1450
for j in range(0,len(lines_cluster_set)):
if abs(lines_cluster_set[j][6]) < 1:
if ((MiddlePoint_heng[0]-(lines_cluster_set[j][2]+lines_cluster_set[j][4])/2)**2 + (MiddlePoint_heng[1]-(lines_cluster_set[j][3]+lines_cluster_set[j][5])/2)**2)**0.5 < distance2point:
distance2point = ((MiddlePoint_heng[0]-(lines_cluster_set[j][2]+lines_cluster_set[j][4])/2)**2 + (MiddlePoint_heng[1]-(lines_cluster_set[j][3]+lines_cluster_set[j][5])/2)**2)**0.5
Point_heng = lines_cluster_set[j]
else:
if ((MiddlePoint_shu[0]-(lines_cluster_set[j][2]+lines_cluster_set[j][4])/2)**2 + (MiddlePoint_shu[1]-(lines_cluster_set[j][3]+lines_cluster_set[j][5])/2)**2)**0.5 < distance2point2:
distance2point2 = ((MiddlePoint_shu[0]-(lines_cluster_set[j][2]+lines_cluster_set[j][4])/2)**2 + (MiddlePoint_shu[1]-(lines_cluster_set[j][3]+lines_cluster_set[j][5])/2)**2)**0.5
Point_shu = lines_cluster_set[j]
need_stronger = 0
something_missing = 0
#缺省矫正
if Point_shu != []:
if Point_heng == []:
cross_x = 78
cross_y = 1616
something_missing = 1
if len(docu_num) == 9:
something_missing = 0
cross_x = 93
cross_y = 1666
if docu_num[3] == '4':
something_missing = 0
cross_x = 78
cross_y = 1655
if docu_num[3] == '4' and docu_num[5] == '6':
cross_x = 78
cross_y = 1665
else:
cross_x = (Point_shu[3]-Point_shu[6]*Point_shu[2]-Point_heng[3]+Point_heng[6]*Point_heng[2])/(Point_heng[6]-Point_shu[6])
cross_y = Point_heng[6]*cross_x + Point_heng[3] - Point_heng[6]*Point_heng[2]
cv.line(image_copy, (Point_heng[2], Point_heng[3]), (Point_heng[4], Point_heng[5]), (0, 0, 255), 2)
cv.line(image_copy, (Point_shu[2], Point_shu[3]), (Point_shu[4], Point_shu[5]), (0, 0, 255), 2)
else:
cross_x = Point_heng[2]
cross_y = Point_heng[3]
need_stronger = 1
if Point_heng != [] and cross_x > Point_heng[2] and len(docu_num) == 9 and docu_num[6] == '0' and docu_num[7] == '9':
cross_x = 50
cross_y = 1586
if Point_heng != [] and cross_x > Point_heng[2] and len(docu_num) == 9 and docu_num[6] == '1' and docu_num[7] == '8':
cross_x = 50
cross_y = 1616
print 'roi_mean:',roi_mean
if len(docu_num) == 9 and docu_num[7] == '8' and roi_mean < 230:
cross_x = 78
cross_y = 1631
if cross_y - 1648 < 3:
if len(docu_num) == 8 and docu_num[5] == '2' and docu_num[6] == '4':
cross_x = 78
cross_y = 1601
if len(docu_num) == 9 and docu_num[6] == '1' and docu_num[7] == '4':
cross_x = 78
cross_y = 1621
if table_style == 3 and cross_y < 1485:
cross_x = 78
cross_y = 1641
print cross_x,cross_y#当下直角顶点位置，标准位置为78,1581
cv.circle(image_copy, (int(cross_x), int(cross_y)), 3,(255,0,0),3)
cv.rectangle(image_copy,(0,1450),(180,rows),(255,0,0),3)
cv.imwrite('line_detect_possible_demo.jpg', image_copy)
rows, cols, channels = ShrinkTranslationRotation.shape
print rows, cols
affineShrinkTranslation = np.array([[1, 0, int(78 - cross_x)], [0, 1, int(1581 - cross_y)]], np.float32)
#affineShrinkTranslation = np.array([[1, 0, int(78 - 78)], [0, 1, int(1581 - 1581)]], np.float32)
shrinkTwoTimesTranslation = cv.warpAffine(ShrinkTranslationRotation, affineShrinkTranslation, (cols, rows))
image_copy = cv.warpAffine(image_copy, affineShrinkTranslation, (cols, rows))
##### 对201100XXX中左下角表格内有额外竖线的表格进行检测Detect_not_shu_line_in_201100XXX #####
if table_style == 2:
shrinkTwoTimesTranslation_copy_copy = shrinkTwoTimesTranslation.copy()
roi = shrinkTwoTimesTranslation[1350:1548, 125:300]#180
cv.rectangle(image_copy,(125,1350),(300,1548),(255,0,0),3)
gray = cv.cvtColor(roi, cv.COLOR_RGB2GRAY)
edges = cv.Canny(gray, 50, 150, apertureSize=3) # 50,150,3
lines = []
lines = cv.HoughLinesP(edges, 1.0, np.pi / 180, 100, 0, minLineLength=60,maxLineGap=20)#50,10,20
print lines
go_through = 0
try:
if lines == None:
go_through = 0
except:
go_through = 1
if go_through == 1:
pi = 3.1415
theta_total = 0
theta_count = 0
for line in lines:
x1, y1, x2, y2 = line[0]
x1 = x1 + 125
x2 = x2 + 125
y1 = y1 + 1350
y2 = y2 + 1350
rho = np.sqrt((x1 - x2)**2 + (y1 - y2)**2)
theta = math.atan(float(y2 - y1)/float(x2 - x1 + 0.001))
print(rho, theta, x1, y1, x2, y2)
if theta > pi/3 or theta < -pi/3:
table_style = 1
if len(docu_num) == 9 and docu_num[6] == '1' and docu_num[7] == '1':
jiaozheng = 1
cv.line(shrinkTwoTimesTranslation_copy_copy, (x1, y1), (x2, y2), (255, 0, 0), 2)
cv.imwrite('shrinkTwoTimesTranslation_copy_copy.jpg', shrinkTwoTimesTranslation_copy_copy)
这段做的是平移校正操作，其中大部分是为了应对数据集中的各种意外情况，譬如表格形式不同。但对于Github中给出的三幅较为统一的图片，核心的操作只有对旋转校正后的表格左下角进行横线、竖线用HoughLinesP()进行提取，归一后求二者的交点。将交点平移至设定好的坐标即可统筹处理所有表格图片。此处对霍夫变换求出的各条直线分别与设定好两个等分点进行距离求解，选出最靠近等分点的横线、竖线作为归一结果。求得交点坐标后用np.array()函数与cv.warpAffine()函数即可完成平移操作。
横线、竖线的交点求得后在旋转后的表格图像上作图如下line_detect_possible_demo.jpg：
5、分别通过对二值化后的表格用长横条、长竖条内核进行开操作，将表格分别化为全横线与全竖线，叠加后提取交点，即可得到表格中每个矩形的四个顶点
##### 提取表格Table_Out #####
#分别通过对二值化后的表格用长横条、长竖条内核进行开操作，将表格分别化为全横线与全竖线，叠加后提取交点，即可得到表格中每个矩形的四个顶点
shrinkTwoTimesTranslation_gray = cv.cvtColor(shrinkTwoTimesTranslation, cv.COLOR_RGB2GRAY)
th2 = cv.adaptiveThreshold(~shrinkTwoTimesTranslation_gray,255,cv.ADAPTIVE_THRESH_MEAN_C,cv.THRESH_BINARY,15,-2)
cv.imwrite('th2.jpg', th2)
#长横条内核处理
shrinkTwoTimesTranslation_copy = shrinkTwoTimesTranslation.copy()
th2_copy = th2.copy()
scale = 44;#20,50,45,40,44
rows,cols = shrinkTwoTimesTranslation_gray.shape
horizontalsize = cols / scale
horizontalStructure = cv.getStructuringElement(cv.MORPH_RECT, (horizontalsize, 1))
erosion = cv.erode(th2,horizontalStructure,iterations = 1)
dilation = cv.dilate(erosion,horizontalStructure,iterations = 1)
#长竖条内核处理
scale = 39;#20,50,45,40,39
horizontalsize2 = rows / scale
horizontalStructure2 = cv.getStructuringElement(cv.MORPH_RECT, (1,horizontalsize2))
erosion2 = cv.erode(th2_copy,horizontalStructure2,iterations = 1)
dilation2 = cv.dilate(erosion2,horizontalStructure2,iterations = 1)
#全横线图与全竖线图叠加，并提取交点
mask = dilation + dilation2
cv.imwrite('mask.jpg', mask)
joints = cv.bitwise_and(dilation, dilation2)
cv.imwrite('joints.jpg', joints)
#根据矩形大小筛选矩形框，并画在矫正后的表格上
#cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE
mask, contours, hierarchy = cv.findContours(mask,cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE)
length = len(contours)
print length
small_rects = []
big_rects = []
for i in range(length):
cnt = contours[i]
area = cv.contourArea(cnt)
if area < 10:
continue
approx = cv.approxPolyDP(cnt, 3, True)#3
x, y, w, h = cv.boundingRect(approx)
rect = (x, y, w, h)
#cv.rectangle(img, (x, y), (x+w, y+h), (0, 0, 255), 3)
roi = joints[y:y+h, x:x+w]
roi, joints_contours, joints_hierarchy = cv.findContours(roi,cv.RETR_LIST,cv.CHAIN_APPROX_SIMPLE)
#print len(joints_contours)
#if h < 80 and h > 20 and w > 10 and len(joints_contours)<=4:
if h < 80 and h > 20 and w > 10 and len(joints_contours)<=6:#important
cv.rectangle(image_copy, (x, y), (x+w, y+h), (255-h*3, h*3, 0), 3)
small_rects.append(rect)
cv.imwrite('table_out.jpg', image_copy)此段是功能实现的核心操作：通过对校正图像进行横向、纵向的投影，并提取直线、直线交点，提取轮廓后通过所需表格框的尺寸数据对交点进行筛选、匹配，即可得到表格中的各个框四个顶点的坐标。此处理对表格中的小框效果拔群，对大型框进行提取可能遇到一框中有多个小框的情况，需要后续加以限制（统计该框内是否有别的顶点）。具体操作是：
对未涂花的原始图像进行灰度变化和自适应阈值的二值化操作adaptiveThreshold()，这样能最大可能地保留表格中的直线，效果如下图th2.jpg：
之后对图像进行先腐蚀erode()后膨胀dilate()的闭操作，腐蚀与膨胀分别使用N*1与1*N的形态学操作中的核进行处理。该操作可以理解为把表格中所有的像素进行横向、纵向的投影，且投影会往原先已存在直线上偏移，而文字状的不构成直线的像素则会在腐蚀操作中被抹消。就像从长条形的横、纵栅格中看表格图片一般。将横向、纵向投影好的图片进行叠加即如下图mask.jpg：
表格被神奇地提取出来了！这一步需要注意，一定要是旋转校正过的图像，因为对于一张倾斜的图像，投影操作会让图像变“正”了，实际上图像并没有被旋转过来，这对后续的坐标提取是致命的。既然已经提取出了表格框，那我们就可以通过bitwise_and()函数提取出表格框中的各个交点，如下图joints.jpg：
接下来即通过findContours()函数寻找图像轮廓，并用contourArea()函数求出各个轮廓的面积，以对过小的轮廓进行排除。通过approxPolyDP()与boundingRect()函数用矩形去包围各个轮廓，即可得到该表格图片中各个矩形框的位置数据。之后通过再次使用findContours()函数遍历图像轮廓，用尺寸数据对矩形框进行筛选即可得到表格中所需的框的集合。在校正图像上画出这个集合里的所有框，如下图table_out.jpg：
6、矫正后的表格中信息的大致位置各在一定范围内，根据大致位置的坐标点筛选出该表中该信息对应的矩形框具体坐标
if table_style == 1:
request_info = (770, 150, 1150-770, 300-150)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(770,150),(1150,300),(255,0,0),1)#登记表编号0
for j in range(0,len(small_rects)):
(x, y, w, h) = small_rects[j]
if x < 1060 and y < 1130 and x+w > 1060 and y+h > 1130 and something_missing == 1:#特殊情况下表格矩形的提取
x_rem = x
y_rem = y
w_rem = w
h_rem = h
for j in range(0,len(small_rects)):
(x, y, w, h) = small_rects[j]
if x < 700 and y < 370 and x+w > 700 and y+h > 370 :
request_info = (x, y, w, h)
request_info_set.append(request_info)
cv.rectangle(shrinkTwoTimesTranslation,(x,y),(x+w,y+h),(255,0,0),1)#性别1
########### 识别 性别1 ###########
if table_style == 1:
x_com = 700#信息所在的大致位置
y_com = 370
if table_style == 2:
x_com = 700
y_com = 335
if table_style == 3:
x_com = 724
y_com = 316
if table_style == 4:
x_com = 724
y_com = 292
if need_stronger == 1:#特殊情况处理
x = 673
y = 385
w = 740 - 673
h = 420 - 385
get_all = 1
cha_num = 3
is_ABO = 0
xingbie_result = ''
return_change, xingbie_result_set = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
print xingbie_result_set
if xingbie_result_set[0] == 2206 or xingbie_result_set[1] == 2206 or xingbie_result_set[2] == 2206:
xingbie_result = '男'
if xingbie_result_set[0] == 775 or xingbie_result_set[1] == 775 or xingbie_result_set[2] == 775:
xingbie_result = '女'
if xingbie_result == '':
xingbie_result = '男'
print "性别: ",xingbie_result
else:
for j in range(0,len(request_info_set)):
(x, y, w, h) = request_info_set[j]
if x < x_com and y < y_com and x+w > x_com and y+h > y_com :#提取信息所在的矩形框
get_all = 1#对手写汉字模型的三个识别结果都进行分析
cha_num = 3#字段预计最长长度
is_ABO = 0#不是血型数据
xingbie_result = ''#初始化
#调用chinese_out.py中的chinese_o()函数进行手写汉字识别
return_change, xingbie_result_set = chinese_out.chinese_o(th2_copy, shrinkTwoTimesTranslation_copy, x, y, w, h, cha_num, get_all, is_ABO)
print xingbie_result_set
#如果三个识别结果中有一个'男'，则确定为男性
if xingbie_result_set[0] == 2206 or xingbie_result_set[1] == 2206 or xingbie_result_set[2] == 2206:
xingbie_result = '男'
#如果三个识别结果中有一个'女'，则确定为女性
if xingbie_result_set[0] == 775 or xingbie_result_set[1] == 775 or xingbie_result_set[2] == 775:
xingbie_result = '女'
#特殊情况处理
if xingbie_result == '':
xingbie_result = '男'
print "性别: ",xingbie_result既然框都已经被提取出来了，那不是大功告成了吗？
并没有。因为我们还需要对所需要的位置的框进行提取，先前提取出的框又不附带标签。因此我们此处需要做两步：
首先是选择标准图像中所需的框位置的中心点作为这个框的“ID”，标准图像即先前所说的坐标(78,1581)所在的那张图片。上文代码里即选择（700,370）作为“性别”数据的所在框的大致位置。
其次是通过遍历的方式筛选出这个中心点坐标所处在的矩形框的坐标。因为框较大，因此即使校正的不够精准，这种方法一样可以将目标框精准地框出，只要标准图像中的那个大致位置不偏移得太过严重。