add verify code

verify/localVerifyCode.py

@@ -0,0 +1,75 @@
+# coding: utf-8
+import base64
+import cv2
+import numpy as np
+from keras import models
+
+
+from verify import pretreatment
+from verify.mlearn_for_image import preprocess_input
+
+
+def get_text(img, offset=0):
+    text = pretreatment.get_text(img, offset)
+    text = cv2.cvtColor(text, cv2.COLOR_BGR2GRAY)
+    text = text / 255.0
+    h, w = text.shape
+    text.shape = (1, h, w, 1)
+    return text
+
+
+def base64_to_image(base64_code):
+    # base64解码
+    img_data = base64.b64decode(base64_code)
+    # 转换为np数组
+    img_array = np.fromstring(img_data, np.uint8)
+    # 转换成opencv可用格式
+    img = cv2.imdecode(img_array, cv2.COLOR_RGB2BGR)
+
+    return img
+
+
+def verify(fn):
+    verify_titles = ['打字机', '调色板', '跑步机', '毛线', '老虎', '安全帽', '沙包', '盘子', '本子', '药片', '双面胶', '龙舟', '红酒', '拖把', '卷尺', '海苔', '红豆', '黑板', '热水袋', '烛台', '钟表', '路灯', '沙拉', '海报', '公交卡', '樱桃', '创可贴', '牌坊', '苍蝇拍', '高压锅', '电线', '网球拍', '海鸥', '风铃', '订书机', '冰箱', '话梅', '排风机', '锅铲', '绿豆', '航母', '电子秤', '红枣', '金字塔', '鞭炮', '菠萝', '开瓶器', '电饭煲', '仪表盘', '棉棒', '篮球', '狮子', '蚂蚁', '蜡烛', '茶盅', '印章', '茶几', '啤酒', '档案袋', '挂钟', '刺绣', '铃铛', '护腕', '手掌印', '锦旗', '文具盒', '辣椒酱', '耳塞', '中国结', '蜥蜴', '剪纸', '漏斗', '锣', '蒸笼', '珊瑚', '雨靴', '薯条', '蜜蜂', '日历', '口哨']
+    # 读取并预处理验证码
+    img = base64_to_image(fn)
+    text = get_text(img)
+    imgs = np.array(list(pretreatment._get_imgs(img)))
+    imgs = preprocess_input(imgs)
+    text_list = []
+    # 识别文字
+    model = models.load_model('model.v2.0.h5')
+    label = model.predict(text)
+    label = label.argmax()
+    text = verify_titles[label]
+    text_list.append(text)
+    # 获取下一个词
+    # 根据第一个词的长度来定位第二个词的位置
+    if len(text) == 1:
+        offset = 27
+    elif len(text) == 2:
+        offset = 47
+    else:
+        offset = 60
+    text = get_text(img, offset=offset)
+    if text.mean() < 0.95:
+        label = model.predict(text)
+        label = label.argmax()
+        text = verify_titles[label]
+        text_list.append(text)
+    print("题目为{}".format(text_list))
+    # 加载图片分类器
+    model = models.load_model('12306.image.model.h5')
+    labels = model.predict(imgs)
+    labels = labels.argmax(axis=1)
+    results = []
+    for pos, label in enumerate(labels):
+        l = verify_titles[label]
+        print(pos+1, l)
+        if l in text_list:
+            results.append(str(pos+1))
+    return results
+
+
+if __name__ == '__main__':
+    verify("verify-img1.jpeg")

verify/mlearn_for_image.py

@@ -0,0 +1,92 @@
+# coding: utf-8
+import sys
+
+import cv2
+import numpy as np
+from keras import models
+from keras import layers
+from keras import optimizers
+from keras.applications import VGG16
+from keras.callbacks import ReduceLROnPlateau
+from keras.preprocessing.image import ImageDataGenerator
+
+
+def preprocess_input(x):
+    x = x.astype('float32')
+    # 我是用cv2来读取的图片，其已经是BGR格式了
+    mean = [103.939, 116.779, 123.68]
+    x -= mean
+    return x
+
+
+def load_data():
+    # 这是统计学专家提供的训练集
+    data = np.load('captcha.npz')
+    train_x, train_y = data['images'], data['labels']
+    train_x = preprocess_input(train_x)
+    # 由于是统计得来的信息，所以在此给定可信度
+    sample_weight = train_y.max(axis=1) / np.sqrt(train_y.sum(axis=1))
+    sample_weight /= sample_weight.mean()
+    train_y = train_y.argmax(axis=1)
+
+    # 这是人工提供的验证集
+    data = np.load('captcha.test.npz')
+    test_x, test_y = data['images'], data['labels']
+    test_x = preprocess_input(test_x)
+    return (train_x, train_y, sample_weight), (test_x, test_y)
+
+
+def learn():
+    (train_x, train_y, sample_weight), (test_x, test_y) = load_data()
+    datagen = ImageDataGenerator(horizontal_flip=True,
+                                 vertical_flip=True)
+    train_generator = datagen.flow(train_x, train_y, sample_weight=sample_weight)
+    base = VGG16(weights='imagenet', include_top=False, input_shape=(None, None, 3))
+    for layer in base.layers[:-4]:
+        layer.trainable = False
+    model = models.Sequential([
+        base,
+        layers.BatchNormalization(),
+        layers.Conv2D(64, (3, 3), activation='relu', padding='same'),
+        layers.GlobalAveragePooling2D(),
+        layers.BatchNormalization(),
+        layers.Dense(64, activation='relu'),
+        layers.BatchNormalization(),
+        layers.Dropout(0.20),
+        layers.Dense(80, activation='softmax')
+    ])
+    model.compile(optimizer=optimizers.RMSprop(lr=1e-5),
+                  loss='sparse_categorical_crossentropy',
+                  metrics=['accuracy'])
+    model.summary()
+    reduce_lr = ReduceLROnPlateau(verbose=1)
+    model.fit_generator(train_generator, epochs=400,
+                        steps_per_epoch=100,
+                        validation_data=(test_x[:800], test_y[:800]),
+                        callbacks=[reduce_lr])
+    result = model.evaluate(test_x, test_y)
+    print(result)
+    model.save('12306.image.model.h5', include_optimizer=False)
+
+
+def predict(imgs):
+    imgs = preprocess_input(imgs)
+    model = models.load_model('12306.image.model.h5')
+    labels = model.predict(imgs)
+    return labels
+
+
+def _predict(fn):
+    imgs = cv2.imread(fn)
+    imgs = cv2.resize(imgs, (67, 67))
+    imgs.shape = (-1, 67, 67, 3)
+    labels = predict(imgs)
+    print(labels.max(axis=1))
+    print(labels.argmax(axis=1))
+
+
+if __name__ == '__main__':
+    if len(sys.argv) >= 2:
+        _predict(sys.argv[1])
+    else:
+        learn()

verify/pretreatment.py

@@ -0,0 +1,100 @@
+#! env python
+# coding: utf-8
+# 功能：对图像进行预处理，将文字部分单独提取出来
+# 并存放到ocr目录下
+# 文件名为原验证码文件的文件名
+import hashlib
+import os
+import pathlib
+
+import cv2
+import numpy as np
+import requests
+import scipy.fftpack
+
+
+PATH = 'imgs'
+
+
+def download_image():
+    # 抓取验证码
+    # 存放到指定path下
+    # 文件名为图像的MD5
+    url = 'https://kyfw.12306.cn/otn/passcodeNew/getPassCodeNew?module=login&rand=sjrand'
+    r = requests.get(url)
+    fn = hashlib.md5(r.content).hexdigest()
+    with open(f'{PATH}/{fn}.jpg', 'wb') as fp:
+        fp.write(r.content)
+
+
+def download_images():
+    pathlib.Path(PATH).mkdir(exist_ok=True)
+    for idx in range(40000):
+        download_image()
+        print(idx)
+
+
+def get_text(img, offset=0):
+    # 得到图像中的文本部分
+    return img[3:22, 120 + offset:177 + offset]
+
+
+def avhash(im):
+    im = cv2.resize(im, (8, 8), interpolation=cv2.INTER_CUBIC)
+    avg = im.mean()
+    im = im > avg
+    im = np.packbits(im)
+    return im
+
+
+def phash(im):
+    im = cv2.resize(im, (32, 32), interpolation=cv2.INTER_CUBIC)
+    im = scipy.fftpack.dct(scipy.fftpack.dct(im, axis=0), axis=1)
+    im = im[:8, :8]
+    med = np.median(im)
+    im = im > med
+    im = np.packbits(im)
+    return im
+
+
+def _get_imgs(img):
+    interval = 5
+    length = 67
+    for x in range(40, img.shape[0] - length, interval + length):
+        for y in range(interval, img.shape[1] - length, interval + length):
+            yield img[x:x + length, y:y + length]
+
+
+def get_imgs(img):
+    imgs = []
+    for img in _get_imgs(img):
+        imgs.append(phash(img))
+    return imgs
+
+
+def pretreat():
+    if not os.path.isdir(PATH):
+        download_images()
+    texts, imgs = [], []
+    for img in os.listdir(PATH):
+        img = os.path.join(PATH, img)
+        img = cv2.imread(img, cv2.IMREAD_GRAYSCALE)
+        texts.append(get_text(img))
+        imgs.append(get_imgs(img))
+    return texts, imgs
+
+
+def load_data(path='data.npz'):
+    if not os.path.isfile(path):
+        texts, imgs = pretreat()
+        np.savez(path, texts=texts, images=imgs)
+    f = np.load(path)
+    return f['texts'], f['images']
+
+
+if __name__ == '__main__':
+    texts, imgs = load_data()
+    print(texts.shape)
+    print(imgs.shape)
+    imgs = imgs.reshape(-1, 8)
+    print(np.unique(imgs, axis=0).shape)