一、前言

本文提供一个简易的demo，用于对目标阵容的高亮显示，避免猪脑过载或慢速思考。基本流程由两部分组成，一是使用tesseract识别出卡牌池的名字列表，与目标阵容进行比较，得到需要的卡牌坐标；二是基于上述坐标，对卡牌高亮

二、环境设置

博主是在windows上运行，搭配Pycharm使用，使用该教程需要进行以下设置：

将英雄联盟客户端语言设置为英文
游戏内设置分辨率为1920*1080，无边框模式
windows桌面，右键->显示设置->缩放与布局->更改文本、应用等项目的大小->100%
安装tesseract，并配置到系统环境变量
以管理员身份运行pycharm

三、卡牌识别

3.1 候选框截取

import pyscreenshot                 # Screenshot tool, PIL friendly
from ocr import get_text_from_image
from pynput import keyboard


def takeScreenshotROI(coordinates):
    # Take a screenshot of the champion zone only
    return pyscreenshot.grab(coordinates)


def get_champions():

    width, height = (1920, 1080)
    left, top, right, bottom = (int(width * 0.25), int(height * 0.96), int(width * 0.77), int(height * 0.99))
    _width = (right-left)/5

    champions_list = []
    for idx in range(5):

        coordinates = (int(left + idx*_width)+10, top, int(left + (idx+1)*_width)-55, bottom)

        roi_example = takeScreenshotROI(coordinates)
        # 用于测试 正式运行时将 以下两行代码 注释
        # roi_example = Image.open('examples/ (9).png')
        # roi_example = roi_example.crop(coordinates)

        champion_name = get_text_from_image(roi_example)
        champions_list.append(champion_name)

    return champions_list

计算左上和右下坐标coordinates，后根据坐标截取屏幕。由于卡池共有五张牌，需要循环五次

3.2 ocr识别

"""
Contains all code related to turning a screenshot into a string
"""

from typing import Any
import cv2
import numpy as np
from PIL import ImageGrab
import pytesseract

def image_grayscale(image: ImageGrab.Image) -> Any:
    """Converts an image to grayscale so OCR has an easier time deciphering characters"""
    return cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)


def image_thresholding(image: ImageGrab.Image) -> Any:
    """Applies thresholding to the image https://docs.opencv.org/4.x/d7/d4d/tutorial_py_thresholding.html"""
    return cv2.threshold(image, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]


def image_array(image: ImageGrab.Image) -> Any:
    """Turns the image into an array"""
    image = np.array(image)
    image = image[..., :3]
    return image


def image_resize(image: int, scale: int) -> Any:
    """Resizes the image using the scale passed in argument two"""
    (width, height) = (image.width * scale, image.height * scale)
    return image.resize((width, height))


def get_text_from_image(image: ImageGrab.Image, whitelist: str = "") -> str:
    """Takes an image and returns the text"""
    resize = image_resize(image, 3)
    array = image_array(resize)
    grayscale = image_grayscale(array)
    thresholding = image_thresholding(grayscale)
    return pytesseract.image_to_string(thresholding,
                                       config=f'--psm 7 -c tessedit_char_whitelist={
      
      whitelist}').strip()

由3.1候选框截取到的图片将作为参数，传入函数get_text_from_image进行识别，得到卡牌名

四、高亮显示

基于第三章的函数，会得到现在卡牌池的名字列表，将该列表与目标阵容进行比较，得到需高亮位置的坐标，后使用Pyqt5显示

from PyQt5.QtWidgets import *
from PyQt5.QtGui import *
from PyQt5.QtCore import Qt
import sys
from PyQt5 import QtCore

class Window(QMainWindow):
    def __init__(self, display_list):
        super(Window, self).__init__(None, QtCore.Qt.WindowStaysOnTopHint)

        # this will hide the title bar
        self.setWindowFlag(Qt.FramelessWindowHint)

        # set the title
        self.setWindowTitle("no title")
        self.setGeometry(500, 940, 950, 30)

        for idx in display_list:
            self.label = QLabel('recommend', self)
            self.label.setStyleSheet("background-color: lightgreen; border: 3px solid green")
            self.label.setGeometry(200*idx, 0, 100, 30)

        # show all the widgets
        self.show()


    def keyPressEvent(self, e):
        if e.key() == Qt.Key_D:
            self.close()


App = QApplication(sys.argv)
# recommend i th champion
win1 = Window([0, 1, 3])
App.exec()

效果预览
在这里插入图片描述

参考资料

[1]TFT-OCR-BOT
[2]TFT-Auto-Buy
[3]TFT_zynp

云顶之奕 辅助 高亮显示