forevan

import time
import pyautogui
import cv2
import numpy as np
from PIL import ImageGrab

average = [0]

focus = cv2.imread('focus.png', 0)
w, h = focus.shape[::-1]
for _ in range(1000):
    time.sleep(1)
    pyautogui.moveTo(650, 700)
    time.sleep(0.4)
    pyautogui.mouseDown()
    time.sleep(0.4)
    pyautogui.mouseUp()
    time.sleep(0.4)
    base_screen = ImageGrab.grab(bbox=(0, 0, 1200, 450))
    base_screen.save('D:/rybnui/basescreen.png')

    img_rgb = cv2.imread('basescreen.png')
    img_gray = cv2.cvtColor(img_rgb, cv2.COLOR_BGR2GRAY)

    res = cv2.matchTemplate(img_gray, focus, cv2.TM_CCOEFF_NORMED)
    loc = np.where( res >= 0.7)

    for i in range(25):
        try :
            clean_screen = ImageGrab.grab(bbox=(x, y, x + w, y + h))
            mean = np.mean(clean_screen)
            diff = average[-1] - mean
            if diff >= 4:
                pyautogui.moveTo(x/2+15, y/2+15)
                print('Курсор: Я на поплавке')
                pyautogui.mouseDown()
                time.sleep(0.4)
                pyautogui.mouseUp()
                break
            average.append(mean)
        except:
            for pt in zip(*loc[::-1]):
                x = int(pt[0])
                y = int(pt[1])
            time.sleep(0.2)
        pyautogui.moveTo(100, 100)
        try:
            del(x)
            del(y)
        except:
            pass
        average = [0, ]
        time.sleed(1)