1

I have this python script that can detect the Number plate out of a car image. The number plate I have gotten is this number plate image after applying some openCV tools such as edge detection and morphological transformation.

The number plate I got: pic

However, the detected plate number I got after using pytesseract OCR library is : Detected Text : HHOHVBBGE

This is the original Number plate: screenprint

how can i get better accuracy output?

G. Sliepen
  • 7,637
  • 1
  • 15
  • 31
Rinchen_Dev
  • 21
  • 2
  • 1
    Possible duplicate of [Image Preprocessing for OCR - Tessaract](https://stackoverflow.com/questions/51688973/image-preprocessing-for-ocr-tessaract) – stovfl Sep 23 '18 at 10:04

1 Answers1

0

Tesseract is good at detecting text when the letters are quiet distinct from one another, so you would probably need to do more image processing before you pass the image into the tesseract engine.

My suggestion is you train your own text detector using SVM,K-means or CNNs and you will get better accuracies provided you have enough data.

Rakshith G B
  • 816
  • 2
  • 8
  • 24