深度学习案例之验证码识别

本项目介绍利用深度学习技术（tensorflow），来识别验证码（4位验证码，具体的验证码的长度可以自己生成，可以在自己进行训练）

程序分为四个部分

1、生成验证码的程序，可生成数字+字母大小写的任意长度验证码

# coding:utf-8
# name:captcha_gen.pyimport random
import numpy as np
from PIL import Image
from captcha.image import ImageCaptchaNUMBER = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
LOW_CASE = ['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']
UP_CASE = ['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']CAPTCHA_LIST = NUMBER
CAPTCHA_LEN = 4         # 验证码长度
CAPTCHA_HEIGHT = 60     # 验证码高度
CAPTCHA_WIDTH = 160     # 验证码宽度def random_captcha_text(char_set=CAPTCHA_LIST, captcha_size=CAPTCHA_LEN):"""随机生成定长字符串:param char_set: 备选字符串列表:param captcha_size: 字符串长度:return: 字符串"""captcha_text = [random.choice(char_set) for _ in range(captcha_size)]return ''.join(captcha_text)def gen_captcha_text_and_image(width=CAPTCHA_WIDTH, height=CAPTCHA_HEIGHT, save=None):"""生成随机验证码:param width: 验证码图片宽度:param height: 验证码图片高度:param save: 是否保存（None）:return: 验证码字符串，验证码图像np数组"""image = ImageCaptcha(width=width, height=height)# 验证码文本captcha_text = random_captcha_text()captcha = image.generate(captcha_text)# 保存if save:image.write(captcha_text, './img/' + captcha_text + '.jpg')captcha_image = Image.open(captcha)# 转化为np数组captcha_image = np.array(captcha_image)return captcha_text, captcha_imageif __name__ == '__main__':t, im = gen_captcha_text_and_image(save=True)print(t, im.shape)      # (60, 160, 3)

2、工具库，用于调用验证码生成程序来生成训练集

# -*- coding:utf-8 -*-
# name: util.pyimport numpy as np
from captcha_gen import gen_captcha_text_and_image
from captcha_gen import CAPTCHA_LIST, CAPTCHA_LEN, CAPTCHA_HEIGHT, CAPTCHA_WIDTHdef convert2gray(img):"""图片转为黑白，3维转1维:param img: np:return:  灰度图的np"""if len(img.shape) > 2:img = np.mean(img, -1)return imgdef text2vec(text, captcha_len=CAPTCHA_LEN, captcha_list=CAPTCHA_LIST):"""验证码文本转为向量 哑编码 方式:param text::param captcha_len::param captcha_list::return: vector 文本对应的向量形式"""text_len = len(text)    # 欲生成验证码的字符长度if text_len > captcha_len:raise ValueError('验证码最长4个字符')vector = np.zeros(captcha_len * len(captcha_list))      # 生成一个一维向量 验证码长度*字符列表长度for i in range(text_len):vector[captcha_list.index(text[i])+i*len(captcha_list)] = 1     # 找到字符对应在字符列表中的下标值+字符列表长度*i 的 一维向量 赋值为 1return vectordef vec2text(vec, captcha_list=CAPTCHA_LIST, captcha_len=CAPTCHA_LEN):"""验证码向量转为文本:param vec::param captcha_list::param captcha_len::return: 向量的字符串形式"""vec_idx = vectext_list = [captcha_list[int(v)] for v in vec_idx]return ''.join(text_list)def wrap_gen_captcha_text_and_image(shape=(CAPTCHA_HEIGHT, CAPTCHA_WIDTH, 3)):"""返回特定shape图片:param shape::return:"""while True:t, im = gen_captcha_text_and_image()if im.shape == shape:return t, imdef get_next_batch(batch_count=60, width=CAPTCHA_WIDTH, height=CAPTCHA_HEIGHT):"""获取训练图片组:param batch_count: default 60:param width: 验证码宽度:param height: 验证码高度:return: batch_x, batch_yc"""batch_x = np.zeros([batch_count, width * height])batch_y = np.zeros([batch_count, CAPTCHA_LEN * len(CAPTCHA_LIST)])for i in range(batch_count):    # 生成对应的训练集text, image = wrap_gen_captcha_text_and_image()image = convert2gray(image)     # 转灰度numpy# 将图片数组一维化 同时将文本也对应在两个二维组的同一行batch_x[i, :] = image.flatten() / 255batch_y[i, :] = text2vec(text)  # 验证码文本的向量形式# 返回该训练批次return batch_x, batch_yif __name__ == '__main__':x, y = get_next_batch(batch_count=1)    # 默认为1用于测试集print(x, y)

3、训练程序，并将准确率超过0.95的模型保存到 ./model/ 文件夹下

# -*- coding:utf-8 -*-
# name: model_train.pyimport tensorflow as tf
from datetime import datetime
from util import get_next_batch
from captcha_gen import CAPTCHA_HEIGHT, CAPTCHA_WIDTH, CAPTCHA_LEN, CAPTCHA_LISTdef weight_variable(shape, w_alpha=0.01):"""初始化权值:param shape::param w_alpha::return:"""initial = w_alpha * tf.random_normal(shape)return tf.Variable(initial)def bias_variable(shape, b_alpha=0.1):"""初始化偏置项:param shape::param b_alpha::return:"""initial = b_alpha * tf.random_normal(shape)return tf.Variable(initial)def conv2d(x, w):"""卷基层 ：局部变量线性组合，步长为1，模式‘SAME’代表卷积后图片尺寸不变，即零边距:param x::param w::return:"""return tf.nn.conv2d(x, w, strides=[1, 1, 1, 1], padding='SAME')def max_pool_2x2(x):"""池化层：max pooling,取出区域内最大值为代表特征， 2x2 的pool，图片尺寸变为1/2:param x::return:"""return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')def cnn_graph(x, keep_prob, size, captcha_list=CAPTCHA_LIST, captcha_len=CAPTCHA_LEN):"""三层卷积神经网络:param x:   训练集 image x:param keep_prob:   神经元利用率:param size:        大小 (高,宽):param captcha_list::param captcha_len::return: y_conv"""# 需要将图片reshape为4维向量image_height, image_width = sizex_image = tf.reshape(x, shape=[-1, image_height, image_width, 1])# 第一层# filter 定义为3x3x1， 输出32个特征, 即32个filterw_conv1 = weight_variable([3, 3, 1, 32])    # 3*3的采样窗口，32个（通道）卷积核从1个平面抽取特征得到32个特征平面b_conv1 = bias_variable([32])h_conv1 = tf.nn.relu(conv2d(x_image, w_conv1) + b_conv1)    # rulu激活函数h_pool1 = max_pool_2x2(h_conv1)     # 池化h_drop1 = tf.nn.dropout(h_pool1, keep_prob)      # dropout 防止过拟合# 第二层w_conv2 = weight_variable([3, 3, 32, 64])b_conv2 = bias_variable([64])h_conv2 = tf.nn.relu(conv2d(h_drop1, w_conv2) + b_conv2)h_pool2 = max_pool_2x2(h_conv2)h_drop2 = tf.nn.dropout(h_pool2, keep_prob)# 第三层w_conv3 = weight_variable([3, 3, 64, 64])b_conv3 = bias_variable([64])h_conv3 = tf.nn.relu(conv2d(h_drop2, w_conv3) + b_conv3)h_pool3 = max_pool_2x2(h_conv3)h_drop3 = tf.nn.dropout(h_pool3, keep_prob)"""原始：60*160图片 第一次卷积后 60*160 第一池化后 30*80*32第二次卷积后 30*80*32 ，第二次池化后 15*40*64第三次卷积后 15*40*64 ，第三次池化后 7.5*20*64 = > 向下取整 7*20*64经过上面操作后得到 64 个 7*20的平面"""# 全连接层image_height = int(h_drop3.shape[1])image_width = int(h_drop3.shape[2])w_fc = weight_variable([image_height*image_width*64, 1024])     # 上一层有64个神经元 全连接层有1024个神经元b_fc = bias_variable([1024])h_drop3_re = tf.reshape(h_drop3, [-1, image_height*image_width*64])h_fc = tf.nn.relu(tf.matmul(h_drop3_re, w_fc) + b_fc)h_drop_fc = tf.nn.dropout(h_fc, keep_prob)# 输出层w_out = weight_variable([1024, len(captcha_list)*captcha_len])b_out = bias_variable([len(captcha_list)*captcha_len])y_conv = tf.matmul(h_drop_fc, w_out) + b_outreturn y_convdef optimize_graph(y, y_conv):"""优化计算图:param y: 正确值:param y_conv:  预测值:return: optimizer"""# 交叉熵代价函数计算loss 注意 logits 输入是在函数内部进行sigmod操作# sigmod_cross适用于每个类别相互独立但不互斥，如图中可以有字母和数字# softmax_cross适用于每个类别独立且排斥的情况，如数字和字母不可以同时出现loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=y, logits=y_conv))# 最小化loss优化 AdaminOptimizer优化optimizer = tf.train.AdamOptimizer(1e-3).minimize(loss)return optimizerdef accuracy_graph(y, y_conv, width=len(CAPTCHA_LIST), height=CAPTCHA_LEN):"""偏差计算图，正确值和预测值，计算准确度:param y: 正确值 标签:param y_conv:  预测值:param width:   验证码预备字符列表长度:param height:  验证码的大小，默认为4:return:    正确率"""# 这里区分了大小写 实际上验证码一般不区分大小写,有四个值，不同于手写体识别# 预测值predict = tf.reshape(y_conv, [-1, height, width])   #max_predict_idx = tf.argmax(predict, 2)# 标签label = tf.reshape(y, [-1, height, width])max_label_idx = tf.argmax(label, 2)correct_p = tf.equal(max_predict_idx, max_label_idx)    # 判断是否相等accuracy = tf.reduce_mean(tf.cast(correct_p, tf.float32))return accuracydef train(height=CAPTCHA_HEIGHT, width=CAPTCHA_WIDTH, y_size=len(CAPTCHA_LIST)*CAPTCHA_LEN):"""cnn训练:param height:  验证码高度:param width:   验证码宽度:param y_size:  验证码预备字符列表长度*验证码长度（默认为4）:return:"""# cnn在图像大小是2的倍数时性能最高, 如果图像大小不是2的倍数，可以在图像边缘补无用像素# 在图像上补2行，下补3行，左补2行，右补2行# np.pad(image,((2,3),(2,2)), 'constant', constant_values=(255,))acc_rate = 0.95     # 预设模型准确率标准# 按照图片大小申请占位符x = tf.placeholder(tf.float32, [None, height * width])y = tf.placeholder(tf.float32, [None, y_size])# 防止过拟合 训练时启用 测试时不启用 神经元使用率keep_prob = tf.placeholder(tf.float32)# cnn 模型y_conv = cnn_graph(x, keep_prob, (height, width))# 优化optimizer = optimize_graph(y, y_conv)# 计算准确率accuracy = accuracy_graph(y, y_conv)# 启动会话.开始训练saver = tf.train.Saver()sess = tf.Session()sess.run(tf.global_variables_initializer())     # 初始化step = 0    # 步数while 1:print(step)batch_x, batch_y = get_next_batch(64)sess.run(optimizer, feed_dict={x: batch_x, y: batch_y, keep_prob: 0.75})# 每训练一百次测试一次if step % 10 == 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: 1.0})print(datetime.now().strftime('%c'), ' step:', step, ' accuracy:', acc)# 准确率满足要求，保存模型if acc > acc_rate:model_path = "./model/captcha.model"saver.save(sess, model_path, global_step=step)acc_rate += 0.01if acc_rate > 0.99:     # 准确率达到99%则退出breakstep += 1sess.close()if __name__ == '__main__':train()

4、测试模型效果：

# -*- coding:utf-8 -*-
# name: model_test.pyimport tensorflow as tf
from model_train import cnn_graph
from captcha_gen import gen_captcha_text_and_image
from util import vec2text, convert2gray
from util import CAPTCHA_LIST, CAPTCHA_WIDTH, CAPTCHA_HEIGHT, CAPTCHA_LEN
from PIL import Imagedef captcha2text(image_list, height=CAPTCHA_HEIGHT, width=CAPTCHA_WIDTH):"""验证码图片转化为文本:param image_list::param height::param width::return:"""x = tf.placeholder(tf.float32, [None, height * width])keep_prob = tf.placeholder(tf.float32)y_conv = cnn_graph(x, keep_prob, (height, width))saver = tf.train.Saver()with tf.Session() as sess:saver.restore(sess, tf.train.latest_checkpoint('model/'))predict = tf.argmax(tf.reshape(y_conv, [-1, CAPTCHA_LEN, len(CAPTCHA_LIST)]), 2)vector_list = sess.run(predict, feed_dict={x: image_list, keep_prob: 1})vector_list = vector_list.tolist()text_list = [vec2text(vector) for vector in vector_list]return text_listif __name__ == '__main__':text, image = gen_captcha_text_and_image()img = Image.fromarray(image)image = convert2gray(image)image = image.flatten() / 255pre_text = captcha2text([image])print("验证码正确值:", text, ' 模型预测值:', pre_text)img.show()