I am trying to develop a license plate recognition system using Labview/vision assistant. Any tech help, example code or suggestions would be greatly appreciated.
Cheers Sgt F.
Royal Electrical and Mechanical Engineers
Hi Sgt F.
The difficulty of this task is going to vary depending on the consistency of the images you will be analyzing.
For instance:
Where will the camera be located relative to the vehicle?
Will the angle of the vehicle be consistent?
What portion of the entire image will be filled by the license plate?
Will the camera be located indoors with constant lighting, or will it be required to work outside 24 / 7?
Will the type of vehicle (and position of license plate) be consistent or are you aiming for the system to work on everything from motor bikes to HGV?
My suggestion would be to obtain as many example images as possible. If you can obtain a range of images the system will have to work with it will allow you to build up the best possible system. If you could post up an example image and answer the above questions it will be easier to offer you more specific advice.
Thanks,
John
Thanks for the reply.
the system will be modelled indoors (on a scaletrix system) so the angle will be consistant and close to the passing vehicle.
i would like the system to be able to distinguish a rectangular numberplate and then perform ocr.
if this is not acheivable then the system will be hugely depandant on the position of the vehicle at the point of image aquisition.
i do not have an image at the mo but will get one within 24hrs.
thanks.
Hi VBCoder ...
I'm trying to do this project for a Persian license plate.
It seems it is very difficult.
Look at the sample Persian license plate :
http://www.ecapic.ir/image/ECA-100507121136.jpg
PLz Help ...
This is what i see can you explain what you want to read from this?
oh ...
I'm so sorry
This is the a sample license plate which must be read :
http://rapidshare.com/files/384559883/AKS715.jpg.html
(I can't attach a Image ! )
(Also I'm not a Arabic man !)
Why?? See "add attachments" at the bottom of your reply.
I am not able to view it still (For me and other maybe able to ).
I know that in the UK and I believe the US that a number plate has microlens to reflect light back to the camera. Using IR light then the only thing visible is the number plate. Once you have the rectangle of that you can start to segment the characters and use OCR. I understand that OCR will not give the right characters but using substitution would be a good way to go. Normally 5 pictures are taken once the rectangle gets to a certain size and then all 5 are processed and either consensus or best score is used to give the result.
Mike