I am working on number plate recognition system, I have so far successful in detecting numbers but in number plate i want to detect english letter only no need for arabic can any one help me in that ?

For example in this image i need to detect to 'R' letter

when i go to detect it detects arabic letters also as well.

Answer 1

[Image Analyst]

You need to do OCR on the image. Use the OCR capability in the Computer Vision System Toolbox. Perhaps you can tell it to ignore Arabic letters.

[Walter Roberson]

By default the ocr() routine detects the so-called Arabic numerals and Latin letters. You have to tell ocr() specifically if you want to be able to detect true Arabic text.

You could train on just the uppercase Latin letters and use ocr() with that.

[Image Analyst]

From the original post I inferred that his "detection" was simply segmentation (like thresholding) and it found Latin and Arabic letters, but that he wanted Latin letters only, and didn't want the Arabic letters. So maybe ocr() would work fine right out of the box and treat Arabic characters as just like any other random, non alphabetic blobs.

[Walter Roberson]

I just tried ocr() on it, and it failed completely. However if you resize the image down to about 1/3 then ocr() can find something... mostly just the big box, nothing useful.

But if you use

lp = imresize( imread('IMG_0618.JPG'), 0.3);
bl = rgb2gray(lp) < 100;

then you could do region detection on that, pull out the sub-images and ocr them (probably).

[Image Analyst]

It's a heavily studied field. So much so that there are now commercial systems available to do license plate recognition, and many police cars have them. You can find the published algorithms here: http://www.visionbib.com/bibliography/motion-f726.html#License%20Plate%20Recognition,%20Extraction,%20Analysis