Implementation of verification code recognition in Python opencv pytesseract

One 、 Environment configuration

need pillow and pytesseract These two libraries ,pip install Just install it .

install pillow -i http://pypi.douban.com/simple --trusted-host pypi.douban.com
pip install pytesseract -i http://pypi.douban.com/simple --trusted-host pypi.douban.com

Install well Tesseract-OCR.exe

pytesseract Library Configuration ： Search for pytesseract.py, Open the .py file , find tesseract_cmd, Change its value to just installed tesseract.exe The path of .

If you like to program, you can add some Q Group 768671345, You can get free learning materials when you join the group , There is also a big guy to answer your questions ！

Two 、 Verification code recognition

Identification verification code , We need to preprocess the image first , Remove lines or noise that may affect the accuracy of recognition , Improve recognition accuracy .

example 1

import cv2 as cv
import pytesseract
from PIL import Image

def recognize_text(image):

Edge preserving filtering Denoise

dst = cv.pyrMeanShiftFiltering(image, sp=10, sr=150)

Grayscale image

gray = cv.cvtColor(dst, cv.COLOR_BGR2GRAY)

Two valued

ret, binary = cv.threshold(gray, 0, 255, cv.THRESH_BINARY_INV | cv.THRESH_OTSU)

Morphological operation corrosion inflation

erode = cv.erode(binary, None, iterations=2)
dilate = cv.dilate(erode, None, iterations=1)
cv.imshow(‘dilate’, dilate)

Logical operations Let the background be white The font is black Easy to identify

cv.bitwise_not(dilate, dilate)
cv.imshow(‘binary-image’, dilate)

distinguish

test_message = Image.fromarray(dilate)
text = pytesseract.image_to_string(test_message)
print(f’ Recognition result ：{text}’)

src = cv.imread(r’./test/044.png’)
cv.imshow(‘input image’, src)
recognize_text(src)
cv.waitKey(0)
cv.destroyAllWindows()

The operation effect is as follows ：

Recognition result ：3n3D

Process finished with exit code 0

example 2

import cv2 as cv
import pytesseract
from PIL import Image

def recognize_text(image):

Edge preserving filtering Denoise

blur =cv.pyrMeanShiftFiltering(image, sp=8, sr=60)
cv.imshow(‘dst’, blur)

Grayscale image

gray = cv.cvtColor(blur, cv.COLOR_BGR2GRAY)

Two valued

ret, binary = cv.threshold(gray, 0, 255, cv.THRESH_BINARY_INV | cv.THRESH_OTSU)
print(f’ Binarization adaptive threshold ：{ret}’)
cv.imshow(‘binary’, binary)

Morphological operation Get structure elements Open operation

kernel = cv.getStructuringElement(cv.MORPH_RECT, (3, 2))
bin1 = cv.morphologyEx(binary, cv.MORPH_OPEN, kernel)
cv.imshow(‘bin1’, bin1)
kernel = cv.getStructuringElement(cv.MORPH_OPEN, (2, 3))
bin2 = cv.morphologyEx(bin1, cv.MORPH_OPEN, kernel)
cv.imshow(‘bin2’, bin2)

Logical operations Let the background be white The font is black Easy to identify

cv.bitwise_not(bin2, bin2)
cv.imshow(‘binary-image’, bin2)

distinguish

test_message = Image.fromarray(bin2)
text = pytesseract.image_to_string(test_message)
print(f’ Recognition result ：{text}’)

src = cv.imread(r’./test/045.png’)
cv.imshow(‘input image’, src)
recognize_text(src)
cv.waitKey(0)
cv.destroyAllWindows()

The operation effect is as follows ：

Binarization adaptive threshold ：181.0
Recognition result ：8A62N1

Process finished with exit code 0

example 3

import cv2 as cv
import pytesseract
from PIL import Image

def recognize_text(image):

Edge preserving filtering Denoise

blur = cv.pyrMeanShiftFiltering(image, sp=8, sr=60)
cv.imshow(‘dst’, blur)

Grayscale image

gray = cv.cvtColor(blur, cv.COLOR_BGR2GRAY)

Two valued Set the threshold Adaptive threshold Yellow 4 You can't extract it

ret, binary = cv.threshold(gray, 185, 255, cv.THRESH_BINARY_INV)
print(f’ Threshold set by binarization ：{ret}’)
cv.imshow(‘binary’, binary)

Logical operations Let the background be white The font is black Easy to identify

cv.bitwise_not(binary, binary)
cv.imshow(‘bg_image’, binary)

distinguish

test_message = Image.fromarray(binary)
text = pytesseract.image_to_string(test_message)
print(f’ Recognition result ：{text}’)

src = cv.imread(r’./test/045.jpg’)
cv.imshow(‘input image’, src)
recognize_text(src)
cv.waitKey(0)
cv.destroyAllWindows()

The operation effect is as follows ：

Threshold set by binarization ：185.0
Recognition result ：7364

Process finished with exit code 0

If you like to program, you can add some Q Group 768671345, You can get free learning materials when you join the group , There is also a big guy to answer your questions ！