python解决滑动验证码方案

1.获取需要滑动的图片

import base64

if safe_query(page, '//*[contains(text(),"向右滑动完成验证")]'):
            target_img = safe_query(page, '//*[@class="verify-sub-block"]/img')
            if not target_img:
                continue
            target_b64 = target_img.get_attribute("src").split(";base64,", 1)[-1]
            target_bytes = base64.b64decode(target_b64.encode())

 

2.获取滑动背景图片

background_img = safe_query(page, '//*[@class="verify-img-panel"]/img')
            if not background_img:
                continue
            background_b64 = background_img.get_attribute("src").split(";base64,", 1)[-1]
            background_bytes = base64.b64decode(background_b64.encode())

 

3.计算滑块与目标位置距离

import cv2
import numpy
import numpy as np
from io import BytesIO

def match_slider_from_bytes(bg_bytes: bytes, slider_bytes: bytes, debug=False) -> int:
    bg_array = np.frombuffer(bg_bytes, np.uint8)
    slider_array = np.frombuffer(slider_bytes, np.uint8)
    bg = cv2.imdecode(bg_array, cv2.IMREAD_GRAYSCALE)
    slider = cv2.imdecode(slider_array, cv2.IMREAD_GRAYSCALE)
    # 边缘检测
    bg_edge = cv2.Canny(bg, 100, 200)
    slider_edge = cv2.Canny(slider, 100, 200)

    # 模板匹配
    result = cv2.matchTemplate(bg_edge, slider_edge, cv2.TM_CCOEFF_NORMED)
    _, max_val, _, max_loc = cv2.minMaxLoc(result)
    return max_loc[0]  # 返回x坐标

distance = match_slider_from_bytes(background_bytes, target_bytes, False)
ii = "距离:{}".format(distance)
print(ii)

• 核心流程：字节数据→NumPy 数组→灰度图→边缘检测→模板匹配→提取 X 坐标，全程围绕 “精准找到滑块位置” 展开。
• 关键优化：转灰度图 + 边缘检测，剔除了颜色、亮度等干扰因素，让模板匹配只关注滑块的轮廓特征，准确率更高。
• 适用场景：主要用于自动化操作中破解滑块验证码，返回的 X 坐标可直接用于控制滑块拖动的距离。

4.模拟人为滑动

def human_like_drag(page, slider, distance: int):
    box = slider.bounding_box()
    if box is None:
        raise Exception("未能获取滑块坐标")
    start_x = box["x"] + box["width"] / 2
    start_y = box["y"] + box["height"] / 2
    page.mouse.move(start_x, start_y)
    page.mouse.down()
    current_x = start_x
    total_steps = 3
    for i in range(total_steps):
        step = (distance / total_steps) * random.uniform(0.9, 1.1)
        current_x += step
        jitter_y = start_y + random.uniform(-0.5, 0.5)
        page.mouse.move(current_x, jitter_y)
        time.sleep(random.uniform(0.01, 0.03))
    page.mouse.up()


slider_block = safe_query(page, '//*[@class="verify-move-block"]')
 if not slider_block:
       continue
human_like_drag(page, slider_block, distance)
    

• 核心设计思路：通过 “中心点击 + 分步滑动 + 随机步长 + Y 轴抖动 + 微小停顿”，模拟真人操作特征，避免被反爬风控识别。
• 关键细节：所有随机值（步长、抖动、停顿）都是为了打破机器的 “规律性”，贴近人类操作的 “不规整性”。
• 依赖环境：需要导入random和time库，且page对象需是 Playwright/Puppeteer 等支持mouse操作的自动化对象。

5.如果不成功，整个步骤可以循环几次