-2

I have been developing an application for License Plate Recognition and I am able to detect the number plate successfully from the image using OpenCV. But the problem is with the output of Tesseract e.g. if there is a number say "BI6 COL" in image, Tesseract detect it as "BISWCDL" I have tried several image preprocessing methods (resampling,morphological operations, binarization...) but the accuracy in results is still low.

Please suggest how can I improve the quality of output

Roop
  • 11
  • 2

1 Answers1

0

If you are working on number plate recognition, I would recommend you to train Tesseract exclusively for number plate characters as per your country, considering most countries have different fonts used for writing on the number plate. Pre-processing the image will not much of a difference unless you train Tesseract to recognize characters on number plate.

USEFUL LINKS

  • I would recommend you to go through this Stack Overflow thread.
  • If you wish to train Tesseract, check out this git repository.
Community
  • 1
  • 1
bikz05
  • 1,575
  • 12
  • 17