基于TensorFlow的字符验证码破解

背景：

基于生成的字符验证码数据集训练模型，从而实现字符验证码的破解。

代码：

验证码生成方式：

from captcha.image import ImageCaptcha  # pip install captcha
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image
import random
#验证码中的字符, 就不用汉字了
number = ['0','1','2','3','4','5','6','7','8','9']
alphabet = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z']
ALPHABET = ['A','B','C','D','E','F','G','H','I','J','K','L','M','N','O','P','Q','R','S','T','U','V','W','X','Y','Z']
# 验证码一般都无视大小写；验证码长度4个字符
def random_captcha_text(char_set=number+alphabet+ALPHABET, captcha_size=4):
    captcha_text = []
    for i in range(captcha_size):
        c = random.choice(char_set)
        captcha_text.append(c)
    return captcha_text
# 生成字符对应的验证码
def gen_captcha_text_and_image():
    image = ImageCaptcha()
    captcha_text = random_captcha_text()
    captcha_text = ''.join(captcha_text)
    captcha = image.generate(captcha_text)
    #image.write(captcha_text, captcha_text + '.jpg')  # 写到文件
    captcha_image = Image.open(captcha)
    captcha_image = np.array(captcha_image)
    return captcha_text, captcha_image
if __name__ == '__main__':
    # 测试
    text, image = gen_captcha_text_and_image()
    f = plt.figure()
    ax = f.add_subplot(111)
    ax.text(0.1, 0.9,text, ha='center', va='center', transform=ax.transAxes)
    plt.imshow(image)
    plt.show()

注意：如果出现以下错误:
self.draw.draw_arc(xy, start, end, ink) TypeError: must be sequence of length 4, not 2
说明是版本太低，需要更新版本、如果线更新太慢（sudo pip install Pillow –upgrade），可以采用离线方式：
先下载安装包，再sudo pip install /home/jeremy/Pillow-3.3.0.tar.gz

训练：

from gen_captcha import gen_captcha_text_and_image
from gen_captcha import number
from gen_captcha import alphabet
from gen_captcha import ALPHABET
import numpy as np
import tensorflow as tf
text, image = gen_captcha_text_and_image()
print("验证码图像channel:", image.shape)  # (60, 160, 3)
# 图像大小
IMAGE_HEIGHT = 60
IMAGE_WIDTH = 160
MAX_CAPTCHA = len(text)
print("验证码文本最长字符数", MAX_CAPTCHA)   # 验证码最长4字符; 我全部固定为4,可以不固定. 如果验证码长度小于4，用'_'补齐
# 把彩色图像转为灰度图像（色彩对识别验证码没有什么用）
def convert2gray(img):
    if len(img.shape) > 2:
        gray = np.mean(img, -1)
        # 上面的转法较快，正规转法如下
        # r, g, b = img[:,:,0], img[:,:,1], img[:,:,2]
        # gray = 0.2989 * r + 0.5870 * g + 0.1140 * b
        return gray
    else:
        return img
"""
cnn在图像大小是2的倍数时性能最高, 如果你用的图像大小不是2的倍数，可以在图像边缘补无用像素。
np.pad(image,((2,3),(2,2)), 'constant', constant_values=(255,))  # 在图像上补2行，下补3行，左补2行，右补2行
"""
# 文本转向量
char_set = number + alphabet + ALPHABET + ['_']  # 如果验证码长度小于4, '_'用来补齐
CHAR_SET_LEN = len(char_set)
def text2vec(text):
    text_len = len(text)
    if text_len > MAX_CAPTCHA:
        raise ValueError('验证码最长4个字符')
    vector = np.zeros(MAX_CAPTCHA*CHAR_SET_LEN)
    def char2pos(c):
        if c =='_':
            k = 62
            return k
        k = ord(c)-48
        if k > 9:
            k = ord(c) - 55
            if k > 35:
                k = ord(c) - 61
                if k > 61:
                    raise ValueError('No Map') 
        return k
    for i, c in enumerate(text):
        idx = i * CHAR_SET_LEN + char2pos(c)
        vector[idx] = 1
    return vector
# 向量转回文本
def vec2text(vec):
    char_pos = vec.nonzero()[0]
    text=[]
    for i, c in enumerate(char_pos):
        char_at_pos = i #c/63
        char_idx = c % CHAR_SET_LEN
        if char_idx < 10:
            char_code = char_idx + ord('0')
        elif char_idx <36:
            char_code = char_idx - 10 + ord('A')
        elif char_idx < 62:
            char_code = char_idx-  36 + ord('a')
        elif char_idx == 62:
            char_code = ord('_')
        else:
            raise ValueError('error')
        text.append(chr(char_code))
    return "".join(text)
"""
#向量（大小MAX_CAPTCHA*CHAR_SET_LEN）用0,1编码 每63个编码一个字符，这样顺利有，字符也有
vec = text2vec("F5Sd")
text = vec2text(vec)
print(text)  # F5Sd
vec = text2vec("SFd5")
text = vec2text(vec)
print(text)  # SFd5
"""
# 生成一个训练batch
def get_next_batch(batch_size=128):
    batch_x = np.zeros([batch_size, IMAGE_HEIGHT*IMAGE_WIDTH])
    batch_y = np.zeros([batch_size, MAX_CAPTCHA*CHAR_SET_LEN])
    # 有时生成图像大小不是(60, 160, 3)
    def wrap_gen_captcha_text_and_image():
        while True:
            text, image = gen_captcha_text_and_image()
            if image.shape == (60, 160, 3):
                return text, image
    for i in range(batch_size):
        text, image = wrap_gen_captcha_text_and_image()
        image = convert2gray(image)
        batch_x[i,:] = image.flatten() / 255 # (image.flatten()-128)/128  mean为0
        batch_y[i,:] = text2vec(text)
    return batch_x, batch_y
####################################################################
X = tf.placeholder(tf.float32, [None, IMAGE_HEIGHT*IMAGE_WIDTH])
Y = tf.placeholder(tf.float32, [None, MAX_CAPTCHA*CHAR_SET_LEN])
keep_prob = tf.placeholder(tf.float32) # dropout
# 定义CNN
def crack_captcha_cnn(w_alpha=0.01, b_alpha=0.1):
    x = tf.reshape(X, shape=[-1, IMAGE_HEIGHT, IMAGE_WIDTH, 1])
    #w_c1_alpha = np.sqrt(2.0/(IMAGE_HEIGHT*IMAGE_WIDTH)) #
    #w_c2_alpha = np.sqrt(2.0/(3*3*32)) 
    #w_c3_alpha = np.sqrt(2.0/(3*3*64)) 
    #w_d1_alpha = np.sqrt(2.0/(8*32*64))
    #out_alpha = np.sqrt(2.0/1024)
    # 3 conv layer
    w_c1 = tf.Variable(w_alpha*tf.random_normal([3, 3, 1, 32]))
    b_c1 = tf.Variable(b_alpha*tf.random_normal([32]))
    conv1 = tf.nn.relu(tf.nn.bias_add(tf.nn.conv2d(x, w_c1, strides=[1, 1, 1, 1], padding='SAME'), b_c1))
    conv1 = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
    conv1 = tf.nn.dropout(conv1, keep_prob)
    w_c2 = tf.Variable(w_alpha*tf.random_normal([3, 3, 32, 64]))
    b_c2 = tf.Variable(b_alpha*tf.random_normal([64]))
    conv2 = tf.nn.relu(tf.nn.bias_add(tf.nn.conv2d(conv1, w_c2, strides=[1, 1, 1, 1], padding='SAME'), b_c2))
    conv2 = tf.nn.max_pool(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
    conv2 = tf.nn.dropout(conv2, keep_prob)
    w_c3 = tf.Variable(w_alpha*tf.random_normal([3, 3, 64, 64]))
    b_c3 = tf.Variable(b_alpha*tf.random_normal([64]))
    conv3 = tf.nn.relu(tf.nn.bias_add(tf.nn.conv2d(conv2, w_c3, strides=[1, 1, 1, 1], padding='SAME'), b_c3))
    conv3 = tf.nn.max_pool(conv3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
    conv3 = tf.nn.dropout(conv3, keep_prob)
    # Fully connected layer
    w_d = tf.Variable(w_alpha*tf.random_normal([8*32*40, 1024]))
    b_d = tf.Variable(b_alpha*tf.random_normal([1024]))
    dense = tf.reshape(conv3, [-1, w_d.get_shape().as_list()[0]])
    dense = tf.nn.relu(tf.add(tf.matmul(dense, w_d), b_d))
    dense = tf.nn.dropout(dense, keep_prob)
    w_out = tf.Variable(w_alpha*tf.random_normal([1024, MAX_CAPTCHA*CHAR_SET_LEN]))
    b_out = tf.Variable(b_alpha*tf.random_normal([MAX_CAPTCHA*CHAR_SET_LEN]))
    out = tf.add(tf.matmul(dense, w_out), b_out)
    #out = tf.nn.softmax(out)
    return out
# 训练
def train_crack_captcha_cnn():
    output = crack_captcha_cnn()
    # loss
    #loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(output, Y))
    loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(output, Y))
        # 最后一层用来分类的softmax和sigmoid有什么不同？
    # optimizer 为了加快训练 learning_rate应该开始大，然后慢慢衰
    optimizer = tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)
    predict = tf.reshape(output, [-1, MAX_CAPTCHA, CHAR_SET_LEN])
    max_idx_p = tf.argmax(predict, 2)
    max_idx_l = tf.argmax(tf.reshape(Y, [-1, MAX_CAPTCHA, CHAR_SET_LEN]), 2)
    correct_pred = tf.equal(max_idx_p, max_idx_l)
    accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
    saver = tf.train.Saver()
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        step = 0
        while True:
            batch_x, batch_y = get_next_batch(64)
            _, loss_ = sess.run([optimizer, loss], feed_dict={X: batch_x, Y: batch_y, keep_prob: 0.75})
            print(step, loss_)
            # 每100 step计算一次准确率
            if step % 100 == 0:
                batch_x_test, batch_y_test = get_next_batch(100)
                acc = sess.run(accuracy, feed_dict={X: batch_x_test, Y: batch_y_test, keep_prob: 1.})
                print(step, acc)
                # 如果准确率大于80%,保存模型,完成训练
                if acc > 0.8:
                    saver.save(sess, "crack_capcha.model", global_step=step)
                    break
            step += 1
train_crack_captcha_cnn()

如果训练的过程中出现AttributeError: ‘module’ object has no attribute ‘global_variables_initializer’的报错，用tf.initialize_all_variables方法替代即可。这是版本的定义不同导致的。所以在使用过程中需要注意不同版本在函数使用上的差异。
CNN需要大量的样本进行训练，由于时间和资源有限，测试时可以只使用数字做为验证码字符集。如果使用数字+大小写字母CNN网络有4*62个输出，只使用数字CNN网络有4*10个输出。

训练结束之后，该目录下生成：

其中的checkpoint，crack_capcha.model-5500和crack_capcha.model-5500.meta是在save操作的时候保存下来的。其中的5500其实是外层迭代的次数。所以在测试的时候，需要加载的模型其实是crack_capcha.model-5500文件。

使用训练的模型识别验证码：

利用上述的训练模型进行验证

def crack_captcha(captcha_image):
    output = crack_captcha_cnn()
    saver = tf.train.Saver()
    with tf.Session() as sess:
        saver.restore(sess, "crack_capcha.model-5500")
        predict = tf.argmax(tf.reshape(output, [-1, MAX_CAPTCHA, CHAR_SET_LEN]), 2)
        text_list = sess.run(predict, feed_dict={X: [captcha_image], keep_prob: 1})
        text = text_list[0].tolist()
        vector = np.zeros(MAX_CAPTCHA*CHAR_SET_LEN)
        i = 0
        for n in text:
                vector[i*CHAR_SET_LEN + n] = 1
                i += 1
        return vec2text(vector)

text, image = gen_captcha_text_and_image()
image = convert2gray(image)
image = image.flatten() / 255
predict_text = crack_captcha(image)
print("正确: {} 预测: {}".format(text, predict_text))

测试验证结果：