Python搭建Keras CNN模型破解网站验证码的实现

# -*- coding: utf-8 -*-
import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
from matplotlib import pyplot as plt
 
from keras.utils import np_utils, plot_model
from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Activation, Flatten
from keras.callbacks import EarlyStopping
from keras.layers import Conv2D, MaxPooling2D
 
# 读取数据
df = pd.read_csv('./data.csv')
 
# 标签值
vals = range(31)
keys = ['1','2','3','4','5','6','7','8','9','A','B','C','D','E','F','G','H','J','K','L','N','P','Q','R','S','T','U','V','X','Y','Z']
label_dict = dict(zip(keys, vals))
 
x_data = df[['v'+str(i+1) for i in range(320)]]
y_data = pd.DataFrame({'label':df['label']})
y_data['class'] = y_data['label'].apply(lambda x: label_dict[x])
 
# 将数据分为训练集和测试集
X_train, X_test, Y_train, Y_test = train_test_split(x_data, y_data['class'], test_size=0.3, random_state=42)
x_train = np.array(X_train).reshape((1167, 20, 16, 1))
x_test = np.array(X_test).reshape((501, 20, 16, 1))
 
# 对标签值进行one-hot encoding
n_classes = 31
y_train = np_utils.to_categorical(Y_train, n_classes)
y_val = np_utils.to_categorical(Y_test, n_classes)
 
input_shape = x_train[0].shape
 
# CNN模型
model = Sequential()
 
# 卷积层和池化层
model.add(Conv2D(32, kernel_size=(3, 3), input_shape=input_shape, padding='same'))
model.add(Activation('relu'))
model.add(Conv2D(32, kernel_size=(3, 3), padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2), padding='same'))
 
# Dropout层
model.add(Dropout(0.25))
 
model.add(Conv2D(64, kernel_size=(3, 3), padding='same'))
model.add(Activation('relu'))
model.add(Conv2D(64, kernel_size=(3, 3), padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2), padding='same'))
 
model.add(Dropout(0.25))
 
model.add(Conv2D(128, kernel_size=(3, 3), padding='same'))
model.add(Activation('relu'))
model.add(Conv2D(128, kernel_size=(3, 3), padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2), padding='same'))
 
model.add(Dropout(0.25))
 
model.add(Flatten())
 
# 全连接层
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(128, activation='relu'))
model.add(Dense(n_classes, activation='softmax'))
 
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
 
# plot model
##plot_model(model, to_file=r'./model.png', show_shapes=True)
 
# 模型训练
callbacks = [EarlyStopping(monitor='val_acc', patience=5, verbose=1)]
batch_size = 64
n_epochs = 100
history = model.fit(x_train, y_train, batch_size=batch_size, epochs=n_epochs, \
          verbose=1, validation_data=(x_test, y_val), callbacks=callbacks)
 
mp = './verifycode_Keras.h5'
model.save(mp)
 
# 绘制验证集上的准确率曲线
val_acc = history.history['val_acc']
plt.plot(range(len(val_acc)), val_acc, label='CNN model')
plt.title('Validation accuracy on verifycode dataset')
plt.xlabel('epochs')
plt.ylabel('accuracy')
plt.legend()
plt.show()

......(忽略之前的输出)
Epoch 22/100
 
 64/1167 [>.............................] - ETA: 3s - loss: 0.0399 - acc: 1.0000
 128/1167 [==>...........................] - ETA: 3s - loss: 0.1195 - acc: 0.9844
 192/1167 [===>..........................] - ETA: 2s - loss: 0.1085 - acc: 0.9792
 256/1167 [=====>........................] - ETA: 2s - loss: 0.1132 - acc: 0.9727
 320/1167 [=======>......................] - ETA: 2s - loss: 0.1045 - acc: 0.9750
 384/1167 [========>.....................] - ETA: 2s - loss: 0.1006 - acc: 0.9740
 448/1167 [==========>...................] - ETA: 2s - loss: 0.1522 - acc: 0.9643
 512/1167 [============>.................] - ETA: 1s - loss: 0.1450 - acc: 0.9648
 576/1167 [=============>................] - ETA: 1s - loss: 0.1368 - acc: 0.9653
 640/1167 [===============>..............] - ETA: 1s - loss: 0.1353 - acc: 0.9641
 704/1167 [=================>............] - ETA: 1s - loss: 0.1280 - acc: 0.9659
 768/1167 [==================>...........] - ETA: 1s - loss: 0.1243 - acc: 0.9674
 832/1167 [====================>.........] - ETA: 0s - loss: 0.1577 - acc: 0.9639
 896/1167 [======================>.......] - ETA: 0s - loss: 0.1488 - acc: 0.9665
 960/1167 [=======================>......] - ETA: 0s - loss: 0.1488 - acc: 0.9656
1024/1167 [=========================>....] - ETA: 0s - loss: 0.1427 - acc: 0.9668
1088/1167 [==========================>...] - ETA: 0s - loss: 0.1435 - acc: 0.9669
1152/1167 [============================>.] - ETA: 0s - loss: 0.1383 - acc: 0.9688
1167/1167 [==============================] - 4s 3ms/step - loss: 0.1380 - acc: 0.9683 - val_loss: 0.0835 - val_acc: 0.9760
Epoch 00022: early stopping

# -*- coding: utf-8 -*-
 
import os
import cv2
import numpy as np
 
def split_picture(imagepath):
 
  # 以灰度模式读取图片
  gray = cv2.imread(imagepath, 0)
 
  # 将图片的边缘变为白色
  height, width = gray.shape
  for i in range(width):
    gray[0, i] = 255
    gray[height-1, i] = 255
  for j in range(height):
    gray[j, 0] = 255
    gray[j, width-1] = 255
 
  # 中值滤波
  blur = cv2.medianBlur(gray, 3) #模板大小3*3
 
  # 二值化
  ret,thresh1 = cv2.threshold(blur, 200, 255, cv2.THRESH_BINARY)
 
  # 提取单个字符
  chars_list = []
  image, contours, hierarchy = cv2.findContours(thresh1, 2, 2)
  for cnt in contours:
    # 最小的外接矩形
    x, y, w, h = cv2.boundingRect(cnt)
    if x != 0 and y != 0 and w*h >= 100:
      chars_list.append((x,y,w,h))
 
  sorted_chars_list = sorted(chars_list, key=lambda x:x[0])
  for i,item in enumerate(sorted_chars_list):
    x, y, w, h = item
    cv2.imwrite('test_verifycode/%d.jpg'%(i+1), thresh1[y:y+h, x:x+w])
 
def remove_edge_picture(imagepath):
 
  image = cv2.imread(imagepath, 0)
  height, width = image.shape
  corner_list = [image[0,0] < 127,
          image[height-1, 0] < 127,
          image[0, width-1]<127,
          image[ height-1, width-1] < 127
          ]
  if sum(corner_list) >= 3:
    os.remove(imagepath)
 
def resplit_with_parts(imagepath, parts):
  image = cv2.imread(imagepath, 0)
  os.remove(imagepath)
  height, width = image.shape
 
  file_name = imagepath.split('/')[-1].split(r'.')[0]
  # 将图片重新分裂成parts部分
  step = width//parts   # 步长
  start = 0       # 起始位置
  for i in range(parts):
    cv2.imwrite('./test_verifycode/%s.jpg'%(file_name+'-'+str(i)), \
          image[:, start:start+step])
    start += step
 
def resplit(imagepath):
 
  image = cv2.imread(imagepath, 0)
  height, width = image.shape
 
  if width >= 64:
    resplit_with_parts(imagepath, 4)
  elif width >= 48:
    resplit_with_parts(imagepath, 3)
  elif width >= 26:
    resplit_with_parts(imagepath, 2)
 
# rename and convert to 16*20 size
def convert(dir, file):
 
  imagepath = dir+'/'+file
  # 读取图片
  image = cv2.imread(imagepath, 0)
  # 二值化
  ret, thresh = cv2.threshold(image, 127, 255, cv2.THRESH_BINARY)
  img = cv2.resize(thresh, (16, 20), interpolation=cv2.INTER_AREA)
  # 保存图片
  cv2.imwrite('%s/%s' % (dir, file), img)
 
# 读取图片的数据，并转化为0-1值
def Read_Data(dir, file):
 
  imagepath = dir+'/'+file
  # 读取图片
  image = cv2.imread(imagepath, 0)
  # 二值化
  ret, thresh = cv2.threshold(image, 127, 255, cv2.THRESH_BINARY)
  # 显示图片
  bin_values = [1 if pixel==255 else 0 for pixel in thresh.ravel()]
 
  return bin_values
 
def predict(VerifyCodePath):
 
  dir = './test_verifycode'
  files = os.listdir(dir)
 
  # 清空原有的文件
  if files:
    for file in files:
      os.remove(dir + '/' + file)
 
  split_picture(VerifyCodePath)
 
  files = os.listdir(dir)
  if not files:
    print('查看的文件夹为空！')
  else:
 
    # 去除噪声图片
    for file in files:
      remove_edge_picture(dir + '/' + file)
 
    # 对黏连图片进行重分割
    for file in os.listdir(dir):
      resplit(dir + '/' + file)
 
    # 将图片统一调整至16*20大小
    for file in os.listdir(dir):
      convert(dir, file)
 
    # 图片中的字符代表的向量
    files = sorted(os.listdir(dir), key=lambda x: x[0])
    table = np.array([Read_Data(dir, file) for file in files]).reshape(-1,20,16,1)
 
    # 模型保存地址
    mp = './verifycode_Keras.h5'
    # 载入模型
    from keras.models import load_model
    cnn = load_model(mp)
    # 模型预测
    y_pred = cnn.predict(table)
    predictions = np.argmax(y_pred, axis=1)
 
    # 标签字典
    keys = range(31)
    vals = ['1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'J', 'K', 'L', 'N',
        'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'X', 'Y', 'Z']
    label_dict = dict(zip(keys, vals))
 
    return ''.join([label_dict[pred] for pred in predictions])
 
def main():
 
  dir = './VerifyCode/'
  correct = 0
  for i, file in enumerate(os.listdir(dir)):
    true_label = file.split('.')[0]
    VerifyCodePath = dir+file
    pred = predict(VerifyCodePath)
 
    if true_label == pred:
      correct += 1
    print(i+1, (true_label, pred), true_label == pred, correct)
 
  total = len(os.listdir(dir))
  print('\n总共图片：%d张\n识别正确：%d张\n识别准确率:%.2f%%.'\
     %(total, correct, correct*100/total))
 
main()