Python+Opencv身份证号码区域提取及识别实现

import cv2 as cv
import numpy as np
import matplotlib.pyplot as plt

# 将身份证号码区域从身份证中提取出
def Extract(op_image, sh_image):

 binary, contours, hierarchy = cv.findContours(op_image,
  cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
 contours.remove(contours[0])
 max_x, max_y, max_w, max_h = cv.boundingRect(contours[0])
 color = (0, 0, 0)
 for c in contours:
 x, y, w, h = cv.boundingRect(c)
 cv.rectangle(op_image, (x, y), (x + w, y + h), color, 1)
 cv.rectangle(sh_image, (x, y), (x + w, y + h), color, 1)
 if max_w < w:
  max_x = x
  max_y = y
  max_w = w
  max_h = h
 cut_img = sh_image[max_y:max_y+max_h, max_x:max_x+max_w]
 cv.imshow("The recognized enlarged image", op_image)
 cv.waitKey(0)
 cv.imshow("The recognized binary image", sh_image)
 cv.waitKey(0)
 return cut_img

# 号码内部区域填充(未继续是用此方法)
def Area_filling(image, kernel):
 # The boundary image
 iterate = np.zeros(image.shape, np.uint8)
 iterate[:, 0] = image[:, 0]
 iterate[:, -1] = image[:, -1]
 iterate[0, :] = image[0, :]
 iterate[-1, :] = image[-1, :]
 while True:
 old_iterate = iterate
 iterate_dilation = cv.dilate(iterate, kernel, iterations=1)
 iterate = cv.bitwise_and(iterate_dilation, image)
 difference = cv.subtract(iterate, old_iterate)
 # if difference is all zeros it will return False
 if not np.any(difference):
  break
 return iterate

# 将身份证号码区域再次切割使得一张图片一位号码
def Segmentation(cut_img, kernel, n):
 #首先进行一次号码内空白填充(效果不佳，放弃)
 #area_img = Area_filling(cut_img, kernel)
 #cv.imshow("area_img", area_img)
 #cv.waitKey(0)
 #dilate = cv.dilate(area_img, kernel, iterations=1)
 #cv.imshow("dilate", dilate)
 #cv.waitKey(0)

 cut_copy = cut_img.copy()
 binary, contours, hierarchy = cv.findContours(cut_copy, cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
 contours.remove(contours[0])
 for c in contours:
 x, y, w, h = cv.boundingRect(c)
 for i in range(h):
  for j in range(w):
  # 把首次用findContours()方法识别的轮廓内区域置黑色
  cut_copy[y + i, x + j] = 0
  # cv.rectangle(cut_copy, (x, y), (x + w, y + h), color, 1)
 cv.imshow("Filled image", cut_copy)
 cv.waitKey(0)

 # 尝试进行分割
 binary, contours, hierarchy = cv.findContours(cut_copy, cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
 #tmp_img = cut_img.copy()
 # 如果识别的轮廓数量不是n+1位(首先是一个整个区域的轮廓，然后是n位号码各自的轮廓，身份证和匹配模板分割均用此方法)
 while len(contours)!=n+1:
 if len(contours) < n+1:
  # 如果提取的轮廓数量小于n+1, 说明可能有两位数被识别到一个轮廓中，做一次闭运算，消除数位之间可能存在的连接部分，然后再次尝试提取
  #cut_copy = cv.dilate(cut_copy, kernel, iterations=1)
  cut_copy = cv.morphologyEx(cut_copy, cv.MORPH_CLOSE, kernel)
  cv.imshow("cut_copy", cut_copy)
  cv.waitKey(0)
  # 再次尝试提取身份证区域的轮廓并将轮廓内区域用黑色覆盖
  binary, contours, hierarchy = cv.findContours(cut_copy, cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
  # 去掉提取出的第一个轮廓(第一个轮廓为整张图片)
  contours.remove(contours[0])
  for c in contours:
  x, y, w, h = cv.boundingRect(c)
  for i in range(h):
   for j in range(w):
   cut_copy[y + i, x + j] = 0
   # cv.rectangle(cut_copy, (x, y), (x + w, y + h), color, 1)
  cv.imshow("Filled image", cut_copy)
  cv.waitKey(0)
  #如果findContours()结果为n，跳出
  if len(contours) == n:
  break

 elif len(contours) > n+1:
  # 如果提取的轮廓数量大于n+1, 说明可能有一位数被识别到两个轮廓中，做一次开运算，增强附近身份证区域部分之间的连接部分，然后再次尝试提取
  #cut_copy = cv.erode(cut_copy, kernel, iterations=1)
  cut_copy = cv.morphologyEx(cut_copy, cv.MORPH_OPEN, kernel2)
  cv.imshow("cut_copy", cut_copy)
  cv.waitKey(0)
  #再次尝试提取身份证区域的轮廓并将轮廓内区域用黑色覆盖
  binary, contours, hierarchy = cv.findContours(cut_copy, cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
  #去掉提取出的第一个轮廓(第一个轮廓为整张图片)
  contours.remove(contours[0])
  for c in contours:
  x, y, w, h = cv.boundingRect(c)
  for i in range(h):
   for j in range(w):
   cut_copy[y + i, x + j] = 0
   # cv.rectangle(cut_copy, (x, y), (x + w, y + h), color, 1)
  #cv.imshow("cut_copy", cut_copy)
  #cv.waitKey(0)
  if len(contours) == n:
  break
 # 上述while()中循环完成后，处理的图像基本满足分割要求，进行最后的提取分割
 binary, contours, hierarchy = cv.findContours(cut_copy, cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
 contours.remove(contours[0])
 color = (0, 0, 0)
 for c in contours:
 x, y, w, h = cv.boundingRect(c)
 for i in range(h):
  for j in range(w):
  cv.rectangle(cut_copy, (x, y), (x + w, y + h), color, 1)
  cv.rectangle(cut_img, (x, y), (x + w, y + h), color, 1)

 cv.imshow("Filled image", cut_copy)
 cv.waitKey(0)
 cv.imshow("cut_img", cut_img)
 cv.waitKey(0)
 #print('number:', len(contours))
 # Returns the result of the split
 return contours
 #return cut_img

# Sort排序方法，先将图像分割，由于分割的先后顺序不是按照从左往右，根据横坐标大小将每位身份证号码图片进行排序
def sort(contours, image):
 tmp_num = []
 x_all = []
 x_sort = []
 for c in contours:
 x, y, w, h = cv.boundingRect(c)
 # 使用x坐标来确定身份证号码图片的顺序，把个图片坐标的x值放入x_sort中
 x_sort.append(x)
 # 建立一个用于索引x坐标的列表
 x_all.append(x)
 tmp_img = image[y+1:y+h-1, x+1:x+w-1]
 tmp_img = cv.resize(tmp_img, (40, 60))
 cv.imshow("Number", tmp_img)
 cv.waitKey(0)
 # 将分割的图片缩小至12乘18像素的大小，标准化同时节约模板匹配的时间
 tmp_img = cv.resize(tmp_img, (12, 18))
 tmp_num.append(tmp_img)
 # 利用x_sort排序，用x_all索引，对身份证号码图片排序
 x_sort.sort()
 num_img = []
 for x in x_sort:
 index = x_all.index(x)
 num_img.append(tmp_num[index])
 # 返回排序后图片列表
 return num_img

# 图像识别方法
def MatchImage(img_num, tplt_num):
 # IDnum用于存储最终的身份证字符串
 IDnum = ''
 # 身份证号码18位
 for i in range(18):
 # 存储最大相似度模板的索引以及最大相似度
 max_index = 0
 max_simil = 0
  # 模板有1~9，0，X共11个
 for j in range(11):
  # 存储身份证号码图片与模板之间的相似度
  simil = 0
  for y in range(18):
  for x in range(12):
   # 如果身份证号码图片与模板之间对应位置像素点相同，simil 值自加1
   if img_num[i][y,x] == tplt_num[j][y,x]:
   simil+=1
  if max_simil < simil:
  max_index = j
  max_simil = simil
  print(str(simil)+' ',end='')
 if max_index < 9:
  IDnum += str(max_index+1)
 elif max_index == 9:
  IDnum += str(0)
 else:
  IDnum += 'X'
 print()
 return IDnum

# 最终效果展示
def display(IDnum, image):
 image = cv.resize(image, (960, 90))
 plt.figure(num='ID_Number')
 plt.subplot(111), plt.imshow(image, cmap='gray'), plt.title(IDnum, fontsize=30), plt.xticks([]), plt.yticks([])
 plt.show()


if __name__ == '__main__':
 # 一共三张做测试用身份证图像
 path = 'IDcard01.jpg'
 #path = 'IDcard02.png'
 #path = 'IDcard.jpg'
 id_card = cv.imread(path, 0)
 cv.imshow('Original image', id_card)
 cv.waitKey(0)
 # 将图像转化成标准大小
 id_card = cv.resize(id_card,(1200, 820))
 cv.imshow('Enlarged original image', id_card)
 cv.waitKey(0)
 # 图像二值化
 ret, binary_img = cv.threshold(id_card, 127, 255, cv.THRESH_BINARY)
 cv.imshow('Binary image', binary_img)
 cv.waitKey(0)

 # RECTANGULAR
 kernel = cv.getStructuringElement(cv.MORPH_RECT, (3, 3))
 # RECTANGULAR
 kernel2 = cv.getStructuringElement(cv.MORPH_DILATE, (5, 5))
 #close_img = cv.morphologyEx(binary_img, cv.MORPH_CLOSE, kernel)
 # The corrosion treatment connects the ID Numbers
 erode = cv.erode(binary_img, kernel, iterations=10)
 cv.imshow('Eroded image', erode)
 cv.waitKey(0)

 cut_img = Extract(erode, binary_img.copy())
 cv.imshow("cut_img", cut_img)
 cv.waitKey(0)

 # 存储最终分割的轮廓
 contours = Segmentation(cut_img, kernel, 18)
 # 对图像进行分割并排序
 img_num = sort(contours, cut_img)

 # 识别用的模板
 tplt_path = '/home/image/Pictures/template.jpg'
 tplt_img = cv.imread(tplt_path, 0)
 #cv.imshow('Template image', tplt_img)
 #cv.waitKey(0)

 ret, binary_tplt = cv.threshold(tplt_img, 127, 255, cv.THRESH_BINARY)
 cv.imshow('Binary template image', binary_tplt)
 cv.waitKey(0)

 # 与身份证相同的分割方式
 contours = Segmentation(binary_tplt, kernel, 11)
 tplt_num = sort(contours, binary_tplt)
 # 最终识别出的身份证号码
 IDnum = MatchImage(img_num, tplt_num)
 print('\nID_Number is:', IDnum)
 # 图片展示
 display(IDnum, cut_img)