版上各位大神好，

小弟最近在研究OCR辨識的相關技術，
自己搜尋版上以及其他網站的教學並嘗試自己試做，
目前只要不管是以實際拍照或者螢幕截圖的方式，
只要遇到「數字」皆會出現無法識別的問題...
有經過二值化、色彩對比、亮度等圖片調整，再進行判讀，
目前嘗試使用pytesseract及easyocr兩種套件進行練習皆會遇到這樣的疑問，

程式碼
圖1

from PIL import Image
import pytesseract, cv2, numpy as np, easyocr, datetime, os, requests
path = "img"
jpg = "{}/original/3333.jpg".format(path)

img = cv2.imread(jpg, cv2.IMREAD_REDUCED_GRAYSCALE_2)
img = img[105:165, 115:600]
contrast, brightness = 500, 0
output = img * (contrast/127 + 1) - contrast + brightness
ret, output = cv2.threshold(output, 127, 255, cv2.THRESH_BINARY_INV)
af_img = "{}/after_treatment/{}.jpg".format(path, datetime.datetime.now().strftime("%Y%m%d"))
cv2.imwrite(af_img, output, [cv2.IMWRITE_JPEG_QUALITY, 80])
reader.readtext(af_img)

圖2

from PIL import Image
import pytesseract, cv2, numpy as np, easyocr, datetime, os, requests
path = "img"
reader = easyocr.Reader(['en'])
jpg = "{}/original/2222.jpg".format(path)
now_date, now_datetime = datetime.datetime.now().strftime("%Y-%m-%d"), datetime.datetime.now().strftime("%H : %M")
def img_ocr(img):
    img = cv2.imread(img, cv2.IMREAD_REDUCED_COLOR_8)
    img = img[230:270, 115:190]
    # 曝光& 亮度調整
    contrast, brightness = 200, 0
    output = img * (contrast/127 + 1) - contrast + brightness
    output = 255-output

    af_img = "{}/after_treatment/{}.jpg".format(path, datetime.datetime.now().strftime("%Y%m%d"))

    cv2.imwrite(af_img, output, [cv2.IMWRITE_JPEG_QUALITY, 80])

    result = reader.readtext(af_img)[0][1].replace(" ","")
    return "{} 度".format(result)

原始圖片
圖1

圖2

處理後
圖1

圖2

※由於圖片內容包含其他資訊不便公開，因此僅供自行剪裁後的圖片。
※程式小白，傷眼抱歉，敬請叮嚀指教！