[FIXED] Python Can't extract numbers from photo Linux

Issue

Here is a hard to decypher image. I started a project where i want to take a photo of the coordinates on google earth pro and show them, for now. It works well but not on all surfaces where it will either mess up numbers or show me gibberish.This is my code.What can I do to improve the number detection?

It works by taking a screenshot of my screen, cropping it and taking out the numbers.(I know it executes infinitely for now it is no problem)

from pynput import keyboard
from PIL import ImageGrab, Image, ImageEnhance, ImageFilter
import pyautogui
import pytesseract
import PIL.ImageOps 

pytesseract.pytesseract.tesseract_cmd = r"/usr/bin/tesseract"

h = 1280
w = 1024

leftc = 0.65*h
topc = 0.98*w
rightc = 0.808*h
bottomc = w-4

def on_press(key) :
    if key == keyboard.Key.shift: # handles if key press is shift
        image = ImageGrab.grab(bbox=(0,0,h,w))
        image = image.crop((leftc, topc, rightc, bottomc))
        image = image.resize((202,16),5)
        image.save('sc.png')
        image_to_text = pytesseract.image_to_string(image,lang='eng')
        print(image_to_text)

def on_release(key) :
    if key == keyboard.Key.shift:
        print()

def get_current_key_input() : 
    with keyboard.Listener(on_press=on_press, on_release=on_release) as listener:
        listener.join()

get_current_key_input()     

Solution

I agree with @furas make sure to read his suggested documentation. On the other hand, let this answer be a tutorial for you to dealing with small images. To make an accurate recognition, you need to:

---

• 1. Upsample
• 2. Center the digits
• 3. Apply simple-thresholding.

---

For small images, upsampling and centering is crucial to make the characters or digits readable by human-eye and recognizable by tesseract. Thresholding will make the features (strokes of the character and digit) available.

---

Up sampling
Centering
Threshold
tesseract	44.429595° lon 26.108718"

---

Code:

import cv2
import pytesseract

img = cv2.imread("dCbPd.png")
gry = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
(h, w) = gry.shape[:2]
gry = cv2.resize(gry, (w*4, h*4))
gry = cv2.copyMakeBorder(gry, 40, 40, 40, 40, cv2.BORDER_CONSTANT, value=255)
thr = cv2.threshold(gry, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]
txt = pytesseract.image_to_string(thr)
print(txt)
print("Pytesseract version: {}".format(pytesseract.get_tesseract_version()))
cv2.imshow("thr", thr)
cv2.waitKey(0)

Answered By - Ahx