tesseract for license plate recognition

[bnaef]

Hi all.

A couple of weeks ago, I posted an inquiry on ocropus to get some feed
back re ocropus usability for a LPR application. I received a fair
amount of very useful feedback.

Largely based on that feedback and some preliminary reading about LPR
systems in general, I launched a license plate recognition
specification project that plans on using tesseract and/or gocr as OCR
engine.

We welcome any input from this group with regards to using tesseract
as LPR OCR engine.

With kind regards
Beat Naef

[SteveP]

LPR sounds definitely doable with tesseract, in my opinion. It may
take a little bit of work. I think it would help to have a little
pattern recognition in a front end to locate the license plate in the
image, crop away the rest, zoom it to where the letters are at least
30 pixels high, good and thick and dark and solid, then pass it to
tesseract. I have used the interface to tesseract from RThomas that
was posted on this site for use in a Windows environment and found it
helpful in that environment.

[kevpatts]

I'm looking to achieve LPR with tesseract also. The main problem I can
forsee is that the picture may have to be rotated and/or scewed before
passing to tesseract. In my application I cannot guarentee that the
licence plate will be pictured head on.

[bnaef]

Steve,
I am sorry for getting back to you so late. I have been pocking around
for feedback in many different locations and I am starting to put the
jigsaw together. Obviously the project/ problem is not exactly trivial
- image grabbing, image pre-processing (area of interest, rotation, de-
skew, re-size, character splitting), ocr, image post-processing
(validation, verification,...).

In fact, a couple of days ago I posted another message on this list
asking for input/ collaboration on a lpr specification project (I
started with a specification framework). Given the complexity, I think
it is prudent to first come out with a spec, then do some proof-of-
concept testing and then start with the implementation.

Would you be interested in participating in such a project? What is
your expertise in OCR and image processing?

With kind regards
Beat

On Jan 30, 2:31 pm, SteveP <SPohor...@sjm.com> wrote:
> LPRsounds definitely doable with tesseract, in my opinion. It may

> take a little bit of work. I think it would help to have a little
> pattern recognition in a front end to locate the license plate in the
> image, crop away the rest, zoom it to where the letters are at least
> 30 pixels high, good and thick and dark and solid, then pass it to
> tesseract. I have used the interface to tesseract from RThomas that
> was posted on this site for use in a Windows environment and found it
> helpful in that environment.
>
> On Jan 28, 8:56 am, bnaef <beat.n...@magicwave-sys.com> wrote:
>
> > Hi all.
>
> > A couple of weeks ago, I posted an inquiry on ocropus to get some feed

> > back re ocropus usability for aLPRapplication. I received a fair

> > amount of very useful feedback.
>
> > Largely based on that feedback and some preliminary reading aboutLPR
> > systems in general, I launched a license plate recognition
> > specification project that plans on using tesseract and/or gocr as OCR
> > engine.
>
> > We welcome any input from this group with regards to using tesseract

> > asLPROCR engine.

[Dimitri]

Hi, I am also interested in finding a solution for my hobby project in
LPR. Is anyone a step further?

On Feb 15, 4:48 pm, bnaef <beat.n...@magicwave-sys.com> wrote:
> Steve,
> I am sorry for getting back to you so late. I have been pocking around
> for feedback in many different locations and I am starting to put the
> jigsaw together. Obviously the project/ problem is not exactly trivial
> - image grabbing, image pre-processing (area of interest, rotation, de-
> skew, re-size, character splitting), ocr, image post-processing
> (validation, verification,...).
>
> In fact, a couple of days ago I posted another message on this list
> asking for input/ collaboration on a lpr specification project (I
> started with a specification framework). Given the complexity, I think
> it is prudent to first come out with a spec, then do some proof-of-
> concept testing and then start with the implementation.
>
> Would you be interested in participating in such a project? What is
> your expertise in OCR and image processing?
>
> With kind regards
> Beat
>
> On Jan 30, 2:31 pm, SteveP <SPohor...@sjm.com> wrote:
>
> > LPRsounds definitely doable with tesseract, in my opinion.  It may
> > take a little bit of work.  I think it would help to have a little

> > pattern recognition in a front end to locate the licenseplatein the