跳转到内容
YesCaptchaYesCaptcha 文档

reCAPTCHA v2 图像识别模拟点击（Selenium）

Markdown 版在 ChatGPT 中打开在 Claude 中打开

这篇教程用 Selenium 4 打开 Google 的 reCAPTCHA 演示页，在真实浏览器里完成图片题：读出题目和图片，交给 reCAPTCHA v2 图像识别（ReCaptchaV2Classification）判断该点哪几格，再模拟点击。和拿 token 的方式相比，这种方式在浏览器里真实完成验证，适合 token 通过率低的网站。

不想写代码可以直接用 浏览器插件，它就是基于同一个接口自动点击的。

准备

Section titled “准备”
终端窗口
pip install requests selenium pillow
export YESCAPTCHA_CLIENT_KEY=你的clientKey

需要安装 Chrome。Selenium 4 会自动下载匹配的 chromedriver。

流程

Section titled “流程”
  1. 点复选框。 复选框在 title="reCAPTCHA" 的 iframe 里，先切换进去，点 #recaptcha-anchor。
  2. 切到题目的 iframe。 题目在另一个 iframe（title 包含 “challenge”）里。
  3. 读题目。 .rc-imageselect-desc-wrapper strong 里加粗的文字就是要找的目标，例如 “traffic lights”。接口要的是题目 ID，用对照表把名称换成 ID（/m/015qff），见 题目 ID。浏览器界面用英文最方便对照。
  4. 取图片。 格子里的图片其实是同一张大图，取 td.rc-imageselect-tile img 的 src 下载下来。9 格（3x3）缩放到 300×300，16 格（4x4）缩放到 450×450，转成 Base64。
  5. 识别。 调用 createTask，类型 ReCaptchaV2Classification，传 image、question。加上 "confidence": 0.5，3x3 图片会一次返回所有符合的格子，不用再处理点完后刷新出来的小图。
  6. 点击。 返回的 objects 是格子序号，从 0 开始、从左到右、从上到下，依次点击对应的 td。
  7. 验证。 点 #recaptcha-verify-button。如果又出了新的一轮题目，回到第 3 步；题目窗口关闭就是通过了。
  8. 提交表单。 回到主页面点提交按钮。

完整代码

Section titled “完整代码”
recaptcha_click_selenium.py
"""Click through a reCAPTCHA v2 image challenge with Selenium 4 and ReCaptchaV2Classification.


Flow: open the page -> click the checkbox -> read the target name and the grid image ->
ask YesCaptcha which tiles to click -> click them -> press Verify. Repeat if a new challenge appears.


Run:  pip install requests selenium pillow
      export YESCAPTCHA_CLIENT_KEY=your_client_key
      python recaptcha_click_selenium.py
Each recognized image is charged (see the price list). reCAPTCHA's page structure can change;
adjust the selectors below if a step stops working.
"""
import base64
import io
import os
import time


import requests
from PIL import Image
from selenium import webdriver
from selenium.webdriver.common.by import By
from selenium.webdriver.support import expected_conditions as EC
from selenium.webdriver.support.wait import WebDriverWait


CLIENT_KEY = os.environ["YESCAPTCHA_CLIENT_KEY"]
PAGE_URL = "https://www.google.com/recaptcha/api2/demo"


# Target names shown in the challenge (English UI) -> question IDs.
# Full list: https://docs.yescaptcha.com/en/tasks/recaptcha-classification/
QUESTION_IDS = {
    "taxis": "/m/0pg52", "bus": "/m/01bjv", "buses": "/m/01bjv", "motorcycles": "/m/04_sv",
    "tractors": "/m/013xlm", "chimneys": "/m/01jk_4", "crosswalks": "/m/014xcs",
    "traffic lights": "/m/015qff", "bicycles": "/m/0199g", "parking meters": "/m/015qbp",
    "cars": "/m/0k4j", "bridges": "/m/015kr", "boats": "/m/019jd", "palm trees": "/m/0cdl1",
    "mountains or hills": "/m/09d_r", "fire hydrants": "/m/01pns0", "a fire hydrant": "/m/01pns0",
    "stairs": "/m/01lynh",
}




def classify(image_b64, question):
    task = {"type": "ReCaptchaV2Classification", "image": image_b64, "question": question, "confidence": 0.5}
    data = requests.post(
        "https://api.yescaptcha.com/createTask",
        json={"clientKey": CLIENT_KEY, "task": task},
        timeout=30,
    ).json()
    if data.get("errorId"):
        raise RuntimeError(f"{data.get('errorCode')}: {data.get('errorDescription')}")
    return data["solution"]  # synchronous: the solution comes back right away




def grid_image_b64(url, size):
    """Download the challenge image and scale it to the standard size (300 for 3x3, 450 for 4x4)."""
    img = Image.open(io.BytesIO(requests.get(url, timeout=30).content)).convert("RGB")
    img = img.resize((size, size))
    buf = io.BytesIO()
    img.save(buf, format="JPEG")
    return base64.b64encode(buf.getvalue()).decode()




driver = webdriver.Chrome()
wait = WebDriverWait(driver, 15)
try:
    driver.get(PAGE_URL)


    # 1. Click the checkbox (inside the "reCAPTCHA" iframe)
    wait.until(EC.frame_to_be_available_and_switch_to_it((By.CSS_SELECTOR, 'iframe[title="reCAPTCHA"]')))
    wait.until(EC.element_to_be_clickable((By.ID, "recaptcha-anchor"))).click()
    driver.switch_to.default_content()


    # 2. Switch to the challenge iframe and solve rounds until it closes
    wait.until(EC.frame_to_be_available_and_switch_to_it(
        (By.XPATH, '//iframe[contains(@title, "challenge")]')))
    for _ in range(5):
        target = wait.until(EC.presence_of_element_located(
            (By.CSS_SELECTOR, ".rc-imageselect-desc-wrapper strong"))).text.strip().lower()
        question = QUESTION_IDS.get(target)
        if not question:
            print("unsupported target, reloading:", target)
            driver.find_element(By.ID, "recaptcha-reload-button").click()
            time.sleep(2)
            continue
        tiles = driver.find_elements(By.CSS_SELECTOR, "td.rc-imageselect-tile")
        size = 450 if len(tiles) == 16 else 300
        url = driver.find_element(By.CSS_SELECTOR, "td.rc-imageselect-tile img").get_attribute("src")
        solution = classify(grid_image_b64(url, size), question)
        for index in solution.get("objects", []):
            tiles[index].click()
            time.sleep(0.3)
        time.sleep(1)
        driver.find_element(By.ID, "recaptcha-verify-button").click()
        time.sleep(3)
        driver.switch_to.default_content()
        frame = driver.find_element(By.XPATH, '//iframe[contains(@title, "challenge")]')
        if not frame.is_displayed():  # challenge closed: passed
            break
        driver.switch_to.frame(frame)


    driver.switch_to.default_content()
    driver.find_element(By.ID, "recaptcha-demo-submit").click()
    time.sleep(2)
    print("passed" if "Verification Success" in driver.page_source else "not passed")
finally:
    driver.quit()

注意

Section titled “注意”
  • reCAPTCHA 的页面结构会更新，代码里的选择器失效时要按实际页面调整。
  • 每次识别一张图都会扣点（价格见 reCAPTCHA v2 图像识别）。点击后验证失败的任务可以在 1 小时内 上报，申请返点。
  • 遇到对照表里没有的题目，可以点刷新按钮换一题，或者联系客服增加支持。