旋轉拖動驗證碼解決方案

前因

曾幾何時，你是否被一個旋轉驗證碼而困擾，沒錯今日主題——旋轉驗證碼。

之前也是被他傷透了心，研究了好幾天的js，想直接通過接口傳輸直接解決驗證碼的，然而我失敗了，不過這一次，他來了他來了，他帶着RotNet走來了。

彩虹屁

RotNet也是我無意間發現的，沒錯時隔了好幾個月，他自己出現在我眼前的。這是他的github：https://github.com/d4nst/RotNet/tree/master，他主要是預測圖像的旋轉角度以校正其方向，庫中包括很全，數據集的下載，訓練，預測全都有，而且最最最重要的是，大神提供了模型，我的天。。。這是什麼神仙，你是孫悟空派來拯救我的吧！兄弟！！！

當然有興趣的同學可以看看他的文章，有具體的思路和網絡實現。還有覺得有用的同學可以星一下他的github

好的，話不多說，先看看我最後的成果吧，

思路和修改

然後因爲在跳出驗證碼的時候一般是直接給出圖片的網址，所以我修改了源文件，用來直接讀取網絡圖片和修整圖片大小來適應網絡，

#utils.py


#在RotNetDataGenerator._get_batches_of_transformed_samples中添加響應代碼
#增加讀取網絡圖片的函數

class RotNetDataGenerator(Iterator):
    def _get_batches_of_transformed_samples(self, index_array):
        # create array to hold the images
        batch_x = np.zeros((len(index_array),) + self.input_shape, dtype='float32')
        # create array to hold the labels
        batch_y = np.zeros(len(index_array), dtype='float32')

        # iterate through the current batch
        for i, j in enumerate(index_array):
            if self.filenames is None:
                image = self.images[j]
            else:
                is_color = int(self.color_mode == 'rgb')
                #修改這這一塊{{{{{{{{{
                image = ImageScale(self.filenames[j]) if self.filenames[j][:4].lower()=="http" else cv2.imread(self.filenames[j], is_color)
                h,w=image.shape[:2]
                if h !=224 or w !=224:
                    image = cv2.resize(image, (224, 224), interpolation=cv2.INTER_CUBIC)
                
                #}}}}}}}}
                if is_color:
                    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

            if self.rotate:
                # get a random angle
                rotation_angle = np.random.randint(360)
            else:
                rotation_angle = 0

            # generate the rotated image
            rotated_image = generate_rotated_image(
                image,
                rotation_angle,
                size=self.input_shape[:2],
                crop_center=self.crop_center,
                crop_largest_rect=self.crop_largest_rect
            )

            # add dimension to account for the channels if the image is greyscale
            if rotated_image.ndim == 2:
                rotated_image = np.expand_dims(rotated_image, axis=2)

            # store the image and label in their corresponding batches
            batch_x[i] = rotated_image
            batch_y[i] = rotation_angle

        if self.one_hot:
            # convert the numerical labels to binary labels
            batch_y = to_categorical(batch_y, 360)
        else:
            batch_y /= 360

        # preprocess input images
        if self.preprocess_func:
            batch_x = self.preprocess_func(batch_x)

        return batch_x, batch_y

def ImageScale(url):
    resp = request.urlopen(url)
    image = np.asarray(bytearray(resp.read()), dtype="uint8")
    image = cv2.imdecode(image, cv2.IMREAD_COLOR)
    return image

預測角度，也是根據他的源碼基礎上做修改的，需要注意的是模型位置和測試圖片的位置需要修改爲你電腦上的文件位置

from __future__ import print_function

import os
import numpy as np

from keras.applications.imagenet_utils import preprocess_input
from keras.models import load_model

from utils import RotNetDataGenerator, angle_error


def process_images(input_path,
                   batch_size=64, crop=True):
    #需要修改模型文件位置
    model = load_model("I:\\pythonProject\\RotNet\\rotnet_models\\rotnet_street_view_resnet50_keras2.hdf5", custom_objects={'angle_error': angle_error}, compile=False)

    extensions = ['.jpg', '.jpeg', '.bmp', '.png']

    if os.path.isfile(input_path) or input_path[:4].lower()=="http":
        image_paths = [input_path]

    else:
        image_paths = [os.path.join(input_path, f)
                       for f in os.listdir(input_path)
                       if os.path.splitext(f)[1].lower() in extensions]


    predictions = model.predict_generator(
        RotNetDataGenerator(
            image_paths,
            input_shape=(224, 224, 3),
            batch_size=batch_size,
            one_hot=True,
            preprocess_func=preprocess_input,
            rotate=False,
            crop_largest_rect=True,
            crop_center=True
        ),
        val_samples=len(image_paths)
    )

    predicted_angles = np.argmax(predictions, axis=1)
    print(predicted_angles)
    return predicted_angles



if __name__ == '__main__':
    #修改測試圖片位置，本地地址，或是網絡圖片地址
    process_images("I:\\pythonProject\\RotNet\\data\\test_examples\\008999_4.jpg")

然後通過分析百度指數的js源碼發現旋轉角度的公式是 angle=o/b*360

即o爲拖動的距離，b=底軸寬-按鈕寬

所以我們需要知道的拖動的距離就是  o=angle*360*b

好的，彙總到一起，就可以了。模擬登錄百度指數，而且支持無頭模式

中間有參考一段這位老哥寫的pyppeteer的拖動，https://blog.csdn.net/qq393912540/article/details/91956136

還有這位老哥的反爬策略

https://github.com/wkunzhi/Python3-Spider/blob/master/%E3%80%90%E6%B7%98%E5%AE%9D%E3%80%91%E8%87%AA%E5%8A%A8%E7%99%BB%E9%99%86/auto_login_pyppeteer.py

import asyncio
from pyppeteer import launch
import random
from correct_rotation_for_angle import process_images


async def page_evaluate(page):
    await page.evaluate(
        '''() =>{ Object.defineProperties(navigator,{ webdriver:{ get: () => false } });window.screen.width=1366; }''')
    await page.evaluate('''() =>{ window.navigator.chrome = { runtime: {}, };}''')
    await page.evaluate('''() =>{ Object.defineProperty(navigator, 'languages', { get: () => ['en-US', 'en'] }); }''')
    await page.evaluate('''() =>{ Object.defineProperty(navigator, 'plugins', { get: () => [1, 2, 3, 4, 5,6], }); }''')


async def main(username, password, width, height):

    browser = await launch({'headless': False,#可以無頭
                            'slowMo':1.3,
                            'userDataDir': './userdata',
                            'args': [
                              f'--window-size={width},{height}'
                              '--disable-extensions',
                              '--hide-scrollbars',
                              '--disable-bundled-ppapi-flash',
                              '--mute-audio',
                              '--no-sandbox',
                              '--disable-setuid-sandbox',
                              '--disable-gpu',
                              '--disable-infobars'
                            ],
                            'dumpio': True
                            })
    page = await browser.newPage()
    # 設置瀏覽器頭部
    await page.setUserAgent("Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/83.0.4103.97 Safari/537.36")

    # 設置瀏覽器大小
    await page.setViewport({'width': width, 'height': height})
    # 注入js，防反爬
    await page_evaluate(page)
    res=await page.goto('http://index.baidu.com/v2/index.html')

    await page.waitFor(2000)
    # 獲取登錄位置的文字，如果是登錄就登錄，不是就使用cookie
    elements = await (await(await page.querySelector('.username-text')).getProperty('textContent')).jsonValue()

    if elements == "登錄":

        await page.click(".username-text")
        await asyncio.sleep(1.6)
        # 填寫用戶名
        await page.type('.pass-text-input-userName', username)
        # 填寫密碼
        await page.hover(".pass-text-input-password")
        await asyncio.sleep(0.5)
        await page.mouse.down()
        await asyncio.sleep(random.random())
        await page.mouse.up()
        # await page.click(".pass-text-input-password")
        await page.type('.pass-text-input-password', password)
        # 點擊登錄
        await page.mouse.move(page.mouse._x+random.randint(50,100), page.mouse._y+random.randint(100,200), options={"step": 3})
        await page.hover(".pass-button-submit")
        await page.mouse.down()
        await asyncio.sleep(random.random())
        await page.mouse.up()
        # await page.click(".pass-button-submit")
        await asyncio.sleep(2)
        rotImg = await page.querySelector('.vcode-spin-img')
        # 如果有驗證碼就去旋轉
        while rotImg:
            img_url=await (await(rotImg).getProperty("src")).jsonValue()
            angle=process_images(img_url)[0]
            bottom_line=await (await(await page.querySelector(".vcode-spin-bottom")).getProperty("offsetWidth")).jsonValue()
            button_line = await (await(await page.querySelector(".vcode-spin-button")).getProperty("offsetWidth")).jsonValue()
            b=bottom_line-button_line
            move_line = angle/360*b
            await try_validation(page,move_line)
            # 停個3秒
            await asyncio.sleep(3)
            rotImg = await page.querySelector('.vcode-spin-img')

        #如果有需要短信驗證碼的彈窗的就費了
        no_in = await page.querySelector(".pass-forceverify-wrapper .forceverify-header-a")
        if no_in:
            print("有短信驗證碼廢了")
            await no_in.click()
    # 停個2秒
    await asyncio.sleep(2)
    cookies = await page.cookies()
    # 無頭模式可以打印一下用戶名看看能不能登錄
    elements = await (await(await page.querySelector('.username-text')).getProperty('textContent')).jsonValue()
    print(elements)
    await browser.close()
    if elements == "登錄":
        return None
    return cookies

async def try_validation(page, distance=308):
    # 將距離拆分成兩段，模擬正常人的行爲
    distance1 = distance - 10
    distance2 = 10
    btn_position = await page.evaluate('''
       () =>{
        return {
         x: document.querySelector('.vcode-spin-button').getBoundingClientRect().x,
         y: document.querySelector('.vcode-spin-button').getBoundingClientRect().y,
         width: document.querySelector('.vcode-spin-button').getBoundingClientRect().width,
         height: document.querySelector('.vcode-spin-button').getBoundingClientRect().height
         }}
        ''')
    x = btn_position['x'] + btn_position['width'] / 2
    y = btn_position['y'] + btn_position['height'] / 2
    # print(btn_position)
    await page.mouse.move(x, y)
    await page.mouse.down()
    await page.mouse.move(x + distance1, y, {'steps': 30})
    await page.waitFor(800)
    await page.mouse.move(x + distance1 + distance2, y, {'steps': 20})
    await page.waitFor(800)
    await page.mouse.up()

def baidu_login(username, password, width, height):
    return asyncio.get_event_loop().run_until_complete(main(username, password, width, height))


if __name__ == "__main__":

    width, height = 1366, 768
    username = '你的賬戶'
    password = '你的密碼'

    cookies = baidu_login(username, password, width, height)
    print(cookies)
    if cookies:
        string_cookies = ""
        for each in cookies:
            string_cookies += f"{each['name']}={each['value']};"

最後 

完整的項目放在https://github.com/ShortCJL/RotateCode，注意：需要把模型下載下來解壓到根目錄

今天的感觸就是，讓我們站在巨人的肩膀上快速成長吧。加油，兄弟！！！