nju_captcha.py

import cv2  
import matplotlib.pyplot as plt
import numpy as np
import easyocr
import requests

URL = 'https://authserver.nju.edu.cn/authserver/captcha.html'

webimg = requests.get(URL).content

# read image from web 
nparr = np.frombuffer(webimg, np.uint8)
img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
plt.imshow(img)
plt.show()

flt = cv2.pyrMeanShiftFiltering(img, sp=10, sr=60)
plt.imshow(flt)
plt.show()

gray = cv2.cvtColor(flt, cv2.COLOR_BGR2GRAY)
# show image in grayscale
plt.imshow(gray, cmap='gray')
plt.show()

ret, binary = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)
# show binary image
plt.imshow(binary, cmap='gray')
plt.show()

# reverse binary image to get text in black
binary = cv2.bitwise_not(binary)
# show reversed binary image
plt.imshow(binary, cmap='gray')
plt.show()

WHITELIST = '0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ'

gray = cv2.cvtColor(binary, cv2.COLOR_GRAY2BGR)
reader = easyocr.Reader(['en'], detector=False)
result = reader.recognize(gray, detail=0, allowlist=WHITELIST)
print(result[0].upper())