Ask Your Question

How do I remove unwanted background edges for my ANPR project

asked 2016-01-13 13:00:46 -0500

Sugosh gravatar image

Here is the binary image of the Canny filtered image. (Gaussian Blur 7x7) image description I just need the License plate or any other text if present in the background. How do I remove the background noise? How do I proceed with the text localization?

Thank you.

edit retag flag offensive close merge delete

2 answers

Sort by » oldest newest most voted

answered 2016-01-14 02:44:26 -0500

pklab gravatar image

updated 2016-01-20 03:20:48 -0500

1st locate the licence plate than proceed with text binarization/recognition. To locate the licence plate you could scan the image with horiz line to detect typical text profile (variation at high frequency).

You could check also:

  • The book Mastering OpenCV with Practical Computer Vision Projects by Packt Publishing 2012. Ch5) Number Plate Recognition using SVM and Neural Networks, by David Escrivá. Here is Code for the book
  • Automatic License Plate Recognition on codeplex. It uses OpenCV
  • OpenALPR Automatic License Plate Recognition library on GitHub
  • JavaANPR. Here there is also a doc: Algorithmic and mathematical principles of automatic number plate recognition systems. Thesis in the pdf format.
edit flag offensive delete link more


Thank you for your reply. I read the source code but I'm confused :(. Can you please provide me a link where I could study this? Thank you very much.

Sugosh gravatar imageSugosh ( 2016-01-16 11:20:31 -0500 )edit

@Sugosh Check my updated answer

pklab gravatar imagepklab ( 2016-01-20 03:22:39 -0500 )edit

Thank you :)

Sugosh gravatar imageSugosh ( 2016-01-21 04:05:27 -0500 )edit

@pklab I'm trying to implement Stroke Width Transform to enhance text detection. I'm stuck at finding gradient(perpendicular) direction at each pixel. The approach I have in mind is to find the next connected pixel and find slope of it, then draw a perpendicular using the inverse slope. Please let me know if there is any other approach to find gradient direction of a pixel or how do I solve my method? I'd really be grateful and thank you in advance.

Sugosh gravatar imageSugosh ( 2016-02-01 11:59:52 -0500 )edit

Without effort, with search engines you will find:

pklab gravatar imagepklab ( 2016-02-02 09:59:52 -0500 )edit

@pklab Thank you :D

Sugosh gravatar imageSugosh ( 2016-02-02 10:02:49 -0500 )edit

answered 2019-01-26 08:12:55 -0500

You can use ready solution:

Recognize license plates from many countries in real time without triggers. Work with different sources: IP cameras, WEB cameras, analog cameras connected through video cards, video files (AVI, MP4) and photos. Control devices (barriers, lights, etc.). Send emails and display alarms in user screen on events. Communicate with GSM device. Send data with POST or GET commands to your web server. Has it‘s own web client server. Much more.

edit flag offensive delete link more


and how does this answer the question ? it's just bragging, only.

berak gravatar imageberak ( 2019-01-26 09:40:25 -0500 )edit

Question Tools



Asked: 2016-01-13 13:00:46 -0500

Seen: 937 times

Last updated: Jan 20 '16