pip install pytesseract opencv-python

img = browser_page.query_selector('img.captcha-img')img_url = img.get_attribute('src')

def verify_captcha(image_path):  # image_path 为图片保存路径    image = cv2.imread(image_path)    # 转换为灰度图像    gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)  # 将图片进行高斯模糊    blurred_image = cv2.GaussianBlur(gray_image, (5, 5), 0)    # 使用二值化提高对比度    _, binary_image = cv2.threshold(blurred_image, 128, 255, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)    # 使用PyTesseract进行OCR    recognized_text = pytesseract.image_to_string(binary_image, lang='eng')    print(recognized_text)

def img_distance(x, y):    # x,y 为滑块图片和源图片的文件内容    # 将图片数据转换为numpy数组并解码rgb格式    img_x = np.frombuffer(x, np.uint8)    img = cv2.imdecode(img_x, cv2.COLOR_RGB2BGR)    # 另一个图片做系统操作    img_y = np.frombuffer(y, np.uint8)    template = cv2.imdecode(img_y , cv2.COLOR_RGB2BGR)  # 滑动滑块图片计算两个图片间的距离    res = cv2.matchTemplate(img, template, cv2.TM_CCORR_NORMED)    value = cv2.minMaxLoc(res)[2][0]    distance = value    return distance

def sliding_deal(page):     # x,y为我们在页面上读到的图片链接并请求到的图片文件内容    distance = img_distance(x, y)  # 获取到验证码dom的边界坐标   box = page.locator(".yidun_control").bounding_box()  # 鼠标移到滑块处并按下  page.locator(".yidun_slider__icon").hover()    page.mouse.down()  # 分解步数，我们将滑动行为分成多次进行以模拟人手动拖动行为    steps = 30      for step in range(1, steps + 1):        progress = step / steps        next_step = distance * progress    # 随机取数，让每一步的距离不均等        mov_x = box['x'] + next_step + random.uniform(-2, 2)        print(next_step)        mov_y = box['y'] + random.uniform(-2, 2)    # 移动鼠标拖动滑块移动        page.mouse.move(mov_x, mov_y)    # 每完成一小步停顿0.02-0.1秒        time.sleep(random.uniform(0.02, 0.1))  #鼠标抬起完成滑动进行验证    page.mouse.up()

def img_verify(page):  #从页面获取图片并进行base64    img = page.query_selector('img.yidun_bg-img')    img_url = img.get_attribute('src')    response = requests.get(img_url)    image_type = response.headers['Content-Type']    image_content = response.content    base64_encoded = b64encode(image_content).decode('utf-8')    data_base64 = f'data:{image_type};base64,{base64_encoded}'  # 通过请求打码平台接口得到对应坐标    url = config_base["DETAYUN_URL"]    key = config_base["DETAYUN_KEY"]    json_data = {        "key": key,        "verify_idf_id": 54,        "img1": data_base64,        "img2": ""    }    """  在响应体可以取到对应的坐标    "code": 200,    "msg": "成功",    "data": {        "res_str": "[(82, 89), (267, 64), (259, 116)]"    }    """  response = requests.post(url, json=json_data)

def click_img():  # 定位到图形边框坐标    div = page.query_selector('div.yidun_bgimg')    box = div.bounding_box()    s_x, s_y = box['x'], box['y']    # 分步模拟鼠标人为移动，步数越多移动越慢    step = 20    tag = 0    for bit_x, bit_y in coordinates:    # 每点完一个位置模拟反应时间        time.sleep(random.uniform(0.5, 1))     # 计算出每一小步需要移动的下一个坐标的位置        if tag > 0:            bit_x, bit_y = bit_x - coordinates[tag - 1][0], bit_y - coordinates[tag - 1][1]            x, y = s_x + coordinates[tag - 1][0], s_y + coordinates[tag - 1][1]        else:            x, y = s_x, s_y        m_x, m_y = bit_x / step, bit_y / step        for i in range(step):      # 控制鼠标移动到对应坐标            page.mouse.move(x + m_x * (i + 1), y + m_y * (i + 1))            # 每一小步后短暂停顿模拟人为拖动            time.sleep(random.uniform(0.05, 0.5))    # 控制鼠标按下抬起模拟鼠标点击        page.mouse.down()        page.mouse.up()        tag = tag + 1    exit_count = exit_count + 1    page.wait_for_timeout(2000)