本文將重點介紹機器視覺的一個分支：文字識別，
介紹如何用一些 Python庫來識別和使用在線圖片中的文字
我們可以很輕松的閱讀圖片里的文字，但是機器閱讀這些圖片就會非常困難，利用這種人類用戶可以正常讀取但是大多數機器人都沒法讀取的圖片，驗證碼 (CAPTCHA)就出現了
將圖像翻譯成文字一般被稱為光學文字識別(Optical Character Recognition, OCR)
下載安裝包，下載訓練數據
https://github.com/tesseract-ocr/tesseract/wiki
https://github.com/tesseract-ocr/tesseract/wiki/Data-Files
?
import pytesseractimport urllib.requestfrom PIL import Imageurl = "https://so.gushiwen.org/RandCode.ashx?"urllib.request.urlretrieve(url,'./captcha.png')def cleanImage(path):image = Image.open(path)# 灰度化image = image.convert('L')# 背景雜點刪除data = image.load()#返回圖片的ndarray數據w,h = image.sizefor i in range(w):for j in range(h):if data[i,j] >150: #閾值，估計的值data[i,j] = 255 #純白else:data[i,j] = 0 #純黑image.save('clean_captcha.png')return imageimage = cleanImage('./captcha.png')image.show()result = pytesseract.image_to_string(image)
#
print(result)古詩文網站使用selenium模擬登陸，識別驗證碼
from selenium import webdriver
import pytesseract
from PIL import Image
import timeurl = 'https://so.gushiwen.org/user/login.aspx?from=http://so.gushiwen.org/user/collect.aspx'def get_captcha():driver = webdriver.Chrome()driver.maximize_window()driver.implicitly_wait(10)driver.get(url)# 元素img = driver.find_element_by_id('imgCode')driver.save_screenshot('./poem.png')image = Image.open('./poem.png')# 左上角坐標loc = img.locationprint(loc)# 圖片寬度高度size = img.sizeprint(size)# 矩形區域# 160,260# 234,291rec = (loc['x']+100 , loc['y']+51, 554, 290)captcha = image.crop(rec)# 保存到文件中captcha.save('./captcha.png')return driverdef recognize_captcha():captcha = Image.open('./captcha.png')gray = captcha.convert('L')data = gray.load()w,h = captcha.sizefor x in range(w):for y in range(h):# 0 ~ 255 0純黑，255純白if data[x,y] < 140:data[x,y] = 0else:data[x,y] = 255code = pytesseract.image_to_string(gray)return codedef login(drive,code):drive.find_element_by_id('email').send_keys('xxxxxxx@qq.com')drive.find_element_by_id('pwd').send_keys('xxxxabc')drive.find_element_by_id('code').send_keys(code)time.sleep(3)drive.find_element_by_id('denglu').click()if __name__ == '__main__':drive = get_captcha()# code驗證碼，有可能出錯code = recognize_captcha()# print('----------------',code)login(drive,code)drive.quit()
)