py-hutun-robot/huitun/captcha_ident.py

import random
import re

import cv2
import requests

import cv2 as cv
import numpy as np
from PIL import Image
from playwright.sync_api import sync_playwright, Page

from util.playwright_util import is_element_present


class CaptchaIdent:
    """
    处理灰豚登录的滑块验证码
    """
    def __init__(self, page):
        self.page = page
        self.frame = self.page.frames[1]

    def start(self):
        for i in range(10):
            self.page.wait_for_timeout(1000)
            start_x, start_y = self.get_slide_block_img_and_start()
            self.get_slide_bg_img()
            distance1, distance2 = self.get_slide_distance(start_x, start_y)
            slide_result = self.move_to_notch(distance1, distance2)
            if not slide_result:
                self.refresh_captcha()
            else:
                return
        raise Exception("滑块验证失败")

    def get_slide_bg_img(self):
        """截取滑动验证码背景图片"""
        if self.frame is not None:
            bg_ele = self.frame.query_selector('.tc-bg-img')
            bg_style = bg_ele.evaluate(
                "element => window.getComputedStyle(element).getPropertyValue('background-image')")
            bg_img = re.search(r'url\("([^"]+)"\)', bg_style).group(1)
            r = requests.get(bg_img)
            with open("./slide_bg.png", "wb") as f:
                f.write(r.content)

            # 调整图片大小，根据style内容将宽度调整为320，高度等比例调整
            img = Image.open("./slide_bg.png")
            bg_img = Image.open("./bg.png")
            ratio = img.width / bg_img.width
            img = img.resize(size=(bg_img.width, int(img.height / ratio)))
            img.save("./slide_bg.png")

    def get_slide_block_img_and_start(self):
        """获取滑块图片以及初始x坐标"""
        print("正在获取滑块图片")

        # 首先保存整个登录背景截图
        self.page.wait_for_timeout(2000)
        slideBg = self.frame.query_selector('#slideBg')
        slideBg.screenshot(path="bg.png")

        # 获取滑动验证码所在的iframe
        captcha_frame = self.frame

        # 获取滑块图片
        # .tc-fg-item对应的有三个元素，一个是目标滑块，一个是滑轨，还有一个是滑轨上的按钮
        for i in range(3):
            slide_block_ele = captcha_frame.locator(".tc-fg-item").nth(i)
            slide_block_style = slide_block_ele.get_attribute("style")

            # 滑轨按钮元素的style值中不包含url字符串
            if "url" not in slide_block_style:
                continue

            # 从元素的style值中分析得出只有目标滑块的top值小于150
            top_value = re.search(r'top: (.+)px;', slide_block_style).groups()[0]
            if float(top_value) > 150:
                continue

            # 获取x坐标
            slide_block_x = float(re.search(r'left: (.+)px; top: ', slide_block_style).groups()[0])
            slide_block_y = float(top_value)

            # 通过滑块位置，从背景图中截取滑块图片  # cropped_image = image.crop((left, top, right, bottom))
            slide_block_rect = slide_block_ele.bounding_box()
            bg = Image.open("bg.png")
            # offset = slide_block_rect["width"] // 5  # 从背景图上截取会混入滑块周围的一些像素点，所以加一个偏移值，截取到滑块内部的图片。
            offset = 5
            slide_block_img = bg.crop((slide_block_x + offset, slide_block_y + offset,
                                       slide_block_x + slide_block_rect["width"] - offset,
                                       slide_block_y + slide_block_rect["height"] - offset))
            slide_block_img.save("slide_block.png")
            return slide_block_x + offset, slide_block_y + offset

    def get_slide_distance(self, start_x, slide_bg_y):
        """获取滑动距离"""
        print("正在获取滑动距离")
        # 通过opencv比较图片，获取缺口位置
        slide_bg_img = cv.imread("./slide_bg.png")
        slide_block_img = cv.imread("./slide_block.png")
        block_height = slide_block_img.shape[0]
        # 裁剪对应坐标图片，缩减搜索范围
        slide_bg_img = slide_bg_img[int(slide_bg_y) - 5: int(slide_bg_y + block_height) + 5, :]
        slide_bg_img = self.set_contrast_brightness(slide_bg_img, 1, 100)
        cv.imwrite("./slide_bg_handled.png", slide_bg_img)

        slide_block_img = self.set_contrast_brightness(slide_block_img, 0.4, 0)
        cv.imwrite("./slide_block_handled.png", slide_block_img)
        result = cv.matchTemplate(slide_block_img, slide_bg_img, cv.TM_CCOEFF_NORMED)
        minVal, maxVal, minLoc, maxLoc = cv.minMaxLoc(result)

        th, tw = slide_block_img.shape[:2]
        tl = maxLoc  # 左上角点的坐标
        br = (tl[0] + tw, tl[1] + th)  # 右下角点的坐标

        cv2.rectangle(slide_bg_img, tl, br, (0, 0, 255), 2)  # 绘制矩形
        cv2.rectangle(slide_bg_img, minLoc, (minLoc[0] + tw, minLoc[1] + th), (0, 255, 255), 2)  # 绘制矩形
        cv2.imwrite("./match_result.png", slide_bg_img)  # 保存在本地

        # 距离
        distance1 = minLoc[0] - start_x
        distance2 = maxLoc[0] - start_x
        return distance1, distance2

    def get_slide_distance2(self, start_x):
        """获取滑动距离"""
        print("正在获取滑动距离")
        # 通过opencv比较图片，获取缺口位置
        slide_bg_img = cv2.imread("./slide_bg.png")
        slide_bg_img = cv2.Canny(slide_bg_img, 100, 200)
        cv2.imwrite("./slide_bg_handle1.png", slide_bg_img)
        slide_bg_img = cv2.cvtColor(slide_bg_img, cv2.COLOR_GRAY2RGB)
        cv2.imwrite("./slide_bg_handle2.png", slide_bg_img)

        slide_block_img = cv.imread("./slide_block.png")
        slide_block_img = cv2.Canny(slide_block_img, 100, 200)

        cv.imwrite("./slide_block_handled1.png", slide_block_img)
        slide_block_img = cv2.cvtColor(slide_block_img, cv2.COLOR_GRAY2RGB)
        cv2.imwrite("./slide_block_handled2.png", slide_block_img)  # 保存在本地
        result = cv.matchTemplate(slide_block_img, slide_bg_img, cv.TM_CCOEFF_NORMED)
        minVal, maxVal, minLoc, maxLoc = cv.minMaxLoc(result)

        th, tw = slide_block_img.shape[:2]
        tl = maxLoc  # 左上角点的坐标
        br = (tl[0] + tw, tl[1] + th)  # 右下角点的坐标
        cv2.rectangle(slide_bg_img, tl, br, (0, 0, 255), 2)  # 绘制矩形
        cv2.imwrite("./match_result.png", slide_bg_img)  # 保存在本地

        # 返回缺口的X坐标
        return maxLoc[0] - start_x, minLoc[0] - start_x

    @staticmethod
    def set_contrast_brightness(frame, contrast_value, brightness_value):
        if not contrast_value:
            contrast_value = 0.0

        if not brightness_value:
            brightness_value = 0

        blank = np.zeros(frame.shape, frame.dtype)
        frame = cv.addWeighted(frame, contrast_value, blank, 1 - contrast_value, brightness_value)
        return frame

    @staticmethod
    def get_tracks(distance):
        """获取移动轨迹"""
        tracks = []  # 移动轨迹
        current = 0  # 当前位移
        mid = distance * 3/4   # 减速阈值
        t = 0.5  # 计算间隔
        v = 1  # 初始速度

        while current < distance:
            if current < mid:
                a = random.randint(5, 10)  # 加速度为正5
            else:
                a = random.randint(-5, -3)  # 加速度为负3

            v0 = v  # 初速度 v0
            v = v0 + a * t  # 当前速度
            move = v0 * t + 1 / 2 * a * t * t  # 移动距离
            current += move
            tracks.append(round(current))

        return tracks

    def move_to_notch(self, distance1, distance2):
        """移动滑轨按钮到缺口处"""
        # 获取滑动验证码所在的iframe
        captcha_iframe = self.frame

        for i in range(2):
            # 获取按钮位置，将鼠标移到上方并按下
            slider_btn_rect = captcha_iframe.get_by_alt_text("slider").bounding_box()
            self.page.mouse.move(slider_btn_rect['x'], slider_btn_rect['y'])
            self.page.mouse.down()

            distance = [distance1, distance2][i]
            if distance <= 0:  # 距离不可能小于等于0
                continue

            print(f"正在进行第{i + 1}次滑动，滑动距离{distance}")
            tracks = self.get_tracks(distance)
            for x in tracks:
                self.page.mouse.move(slider_btn_rect['x'] + x, random.randint(-5, 5) + slider_btn_rect['y'])
            self.page.mouse.move(slider_btn_rect['x'] + tracks[-1] + 5, random.randint(-5, 5) + slider_btn_rect['y'])
            self.page.mouse.move(slider_btn_rect['x'] + tracks[-1] - 5, random.randint(-5, 5) + slider_btn_rect['y'])
            self.page.mouse.up()

            # 滑动结束后等待一段时间
            self.page.wait_for_timeout(2000)

            # 寻找按钮是否还存在，不存在的话表明已通过滑动验证码，存在的话尝试下一个距离
            if not is_element_present(self.page, '.ant-modal-body'):
                print("滑动验证通过")
                return True
        return False

    def refresh_captcha(self):
        """刷新验证码"""
        # 获取滑动验证码所在的iframe
        print("刷新验证码")
        self.frame.query_selector('.tc-action-icon').click()
        self.page.wait_for_timeout(2500)
        self.frame = self.page.frames[1]