如何实时捕获游戏分数 - OpenCV 和 Pytesseract

Question

我想实时捕获在 PC 全屏上玩的游戏的分数和计时器，我有下面的工作代码，但是当单独裁剪时它只能识别分数“0”或计时器“1:13”但不是我想要的完整裁剪图像中的“0 1:13 0”。代码从完整图像生成“45”，这很奇怪（可能由于图像中的线条图形而导致错误识别？），但我希望结果为“0 1:13 0”，帮助！

如何删除完整图像中的图形，以便只有文本，因为我认为这就是它不返回 0 1:13 0 的原因，或者还有其他原因吗？

代码：

import cv2
import numpy as np
import pytesseract

from PIL import ImageGrab
from PIL import Image, ImageOps

while True:
    cap1 = ImageGrab.grab(bbox=(740, 260, 1070,300)).convert('L') 
    cap = ImageOps.invert(cap1)

    text = pytesseract.image_to_string(cap,lang='eng', config='--psm 10 --oem 3 -c tessedit_char_whitelist=0123456789:')

    text = text.strip()

    if len(text) > 0:
        print(text)

    if cv2.waitKey(1) == 27:
        break

cv2.destroyAllWindows()

更新的代码（请查看下面的内容并在我遇到错误时提供帮助）：img = cv2.imread(cap) ^^^^^^^^^^^^^^^ 类型错误：无法将对象转换为“文件名”的“str”：

#!/usr/bin/env python3.8

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

while True:

    cap = ImageGrab.grab(bbox=(740, 260, 1070,300))

    # Read Image and Crop Borders
    img = cv2.imread(cap)

    # Gray Color
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    binary = gray.copy()

    # Thresholding
    th1 = 20
    th2 = 200
    binary[binary<th1]=0
    binary[binary>th2]=0
    binary[binary>=th1]=255

    # Resize as original image is small 
    scale_ = 5 # of original size
    width = int(img.shape[1] * scale_ )
    height = int(img.shape[0] * scale_ )
    dim = (width, height)
    # resize image
    resized = cv2.resize(binary, dim, interpolation = cv2.INTER_CUBIC)

    # Filtering
    filtered = cv2.medianBlur(resized, 17)

    # Morhpology
    # res = cv2.erode(resized, None)

    # IMG Dimensions
    h, w = filtered.shape[:2]

    # Cropping
    s1 = filtered[:,:w//3]
    s2 = filtered[:,2*w//3:]
    t = filtered[:,w//3:2*w//3]

    for i in [s1, t, s2]:
        # OCR
        print(pytesseract.image_to_string(i,  lang='eng',config='--psm 7'))

    # Visualization
    disp = cv2.cvtColor(filtered, cv2.COLOR_GRAY2BGR)

    disp[:, w//3] = (0, 0, 255)
    disp[:, 2*w//3] = (0, 0, 255)

    cv2.namedWindow("output", cv2.WINDOW_NORMAL)
    cv2.imshow("output", disp)

    if cv2.waitKey(1) == 27:
            break

cv2.destroyAllWindows()

Answer 1

您可以像 here 那样在 OCR 之前进行经典处理，此外还有 medianFiltering 以去除盐和纸张噪声，然后将图像分成三分之三以分别检测：

输出

0

1:13

0

#!/usr/bin/env python3.8

import cv2
import numpy as np
import pytesseract

im_path="./"
im_name = "2.jpg"

# Read Image and Crop Borders
img = cv2.imread(im_path+im_name)

# Gray Color
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

binary = gray.copy()

# Thresholding
th1 = 20
th2 = 200
binary[binary<th1]=0
binary[binary>th2]=0
binary[binary>=th1]=255

# Resize as original image is small 
scale_ = 5 # of original size
width = int(img.shape[1] * scale_ )
height = int(img.shape[0] * scale_ )
dim = (width, height)
# resize image
resized = cv2.resize(binary, dim, interpolation = cv2.INTER_CUBIC)

# Filtering
filtered = cv2.medianBlur(resized, 17)

# Morhpology
# res = cv2.erode(resized, None)

# IMG Dimensions
h, w = filtered.shape[:2]

# Cropping
s1 = filtered[:,:w//3]
s2 = filtered[:,2*w//3:]
t = filtered[:,w//3:2*w//3]

for i in [s1, t, s2]:
    # OCR
    print(pytesseract.image_to_string(i,  lang='eng',config='--psm 7'))

# Visualization
disp = cv2.cvtColor(filtered, cv2.COLOR_GRAY2BGR)

disp[:, w//3] = (0, 0, 255)
disp[:, 2*w//3] = (0, 0, 255)

cv2.namedWindow("output", cv2.WINDOW_NORMAL)
cv2.imshow("output", disp)
cv2.waitKey(0)