python识别验证码-PIL+pytesseract

直接上代码

验证码识别目标：http://op2c.ikang.com/login

比如：

 

步骤：

1.图片二值化变色

2.剪裁图片，去掉多余的影响识别的部分，只保留验证码主体

3.去除噪点

4.识别验证码

from PIL import Image, ImageDraw
import pytesseract
import requests
import base64
from time import strftime, localtime

pic_time = strftime('%Y_%m_%d_%H_%M_%S', localtime())


def get_pic(pic_time, url='http://op2c.ikang.com/admin/captcha/image'):
    pic_data = requests.get(url)
    pic_data = pic_data.json()['data']['captcha']
    img = base64.b64decode(pic_data)


    pic_path = 'C:/Users/Administrator/Desktop/验证码/{}.jpg'.format(pic_time)
    
    file = open(pic_path,'wb') 
    file.write(img)  
    file.close()
    print(u'验证码已存储-0')
    return pic_path


def oper_pic(path, pic_time):
    img = Image.open(path)
    box = (55, 15, 195, 60)
    # 截图，只保留有验证码的部分
    img = img.crop(box)
    img = img.convert("RGBA")
    pixdata = img.load()
    ## 二值化
    for y in range(img.size[1]):
        for x in range(img.size[0]):
            if pixdata[x, y][0] < 90:
                pixdata[x, y] = (0, 0, 0, 255)

    for y in range(img.size[1]):
        for x in range(img.size[0]):
            if pixdata[x, y][1] < 136:
                pixdata[x, y] = (0, 0, 0, 255)

    for y in range(img.size[1]):
        for x in range(img.size[0]):
            if pixdata[x, y][2] > 0:
                pixdata[x, y] = (255, 255, 255, 255)

    pic_path = 'C:/Users/Administrator/Desktop/验证码/{}_1.png'.format(pic_time)
    
    img.save(pic_path)
    print(u'验证码已存储-1')
    return img


def recognize_captcha(im):
    num = pytesseract.image_to_string(im)
    return num.replace(' ', '')


res = recognize_captcha(oper_pic(get_pic(pic_time), pic_time))
print(res)

目前这玩意针对上边这种类型的验证码识别准确率在80%左右，勉强能用，先记录下，以后再优化。