Prev: Plotting , axes
Next: Inport structure
From: Mije Jusoh on 20 Jul 2010 05:21
Hi,

I'm currently doing a project on car plate recognition which use template matching and correlation value. I have read lots of other ppl works on this matter and they have proven of achieving high accuracy rate. However, my project is not like the others. I found out that the result is really low where only few of the car plates have been recognized correctly. I would like to know what are the possibilities that these wrong recognition might occur? Was it because of I have prepared a not really good template data? what is the correct way to prepare for the template data? Let say, I have 60 images of car plates with character 'C' on it, and I just took 10 images of it (with variation of sizes). After segmentation process, I managed to isolated the characters into several regions and i've resize all regions into a same size (60x12). Then I accumulate the pixel '1' by column and lets say
that the data of pixel '1' is now in the size of 1x12. The same operations have been applied to all 10 images and I used all 10 data (with size 1x12) to find the average and I used this average data as template image for character 'C'. Is that a proper way to do it? please advise because i'm still new in this area and that is why i'm starting with so-called simple technique.
Thank you very much for any help.
From: us on 20 Jul 2010 05:30
"Mije Jusoh" <mspooh21(a)yahoo.com> wrote in message <i23pq8$933$1(a)fred.mathworks.com>...
> Hi,
>
> I'm currently doing a project on car plate recognition which use template matching and correlation value. I have read lots of other ppl works on this matter and they have proven of achieving high accuracy rate. However, my project is not like the others. I found out that the result is really low where only few of the car plates have been recognized correctly. I would like to know what are the possibilities that these wrong recognition might occur? Was it because of I have prepared a not really good template data? what is the correct way to prepare for the template data? Let say, I have 60 images of car plates with character 'C' on it, and I just took 10 images of it (with variation of sizes). After segmentation process, I managed to isolated the characters into several regions and i've resize all regions into a same size (60x12). Then I accumulate the pixel '1' by column and lets say
> that the data of pixel '1' is now in the size of 1x12. The same operations have been applied to all 10 images and I used all 10 data (with size 1x12) to find the average and I used this average data as template image for character 'C'. Is that a proper way to do it? please advise because i'm still new in this area and that is why i'm starting with so-called simple technique.
> Thank you very much for any help.

nice theoretical framework and great effort...
now: where's your question re ML(?)...

us
From: Steven_Lord on 20 Jul 2010 09:17


"Mije Jusoh" <mspooh21(a)yahoo.com> wrote in message
news:i23pq8$933$1(a)fred.mathworks.com...
> Hi,
>
> I'm currently doing a project on car plate recognition which use template
> matching and correlation value. I have read lots of other ppl works on
> this matter and they have proven of achieving high accuracy rate. However,
> my project is not like the others. I found out that the result is really
> low where only few of the car plates have been recognized correctly. I
> would like to know what are the possibilities that these wrong recognition
> might occur? Was it because of I have prepared a not really good template
> data? what is the correct way to prepare for the template data? Let say, I
> have 60 images of car plates with character 'C' on it, and I just took 10
> images of it (with variation of sizes). After segmentation process, I
> managed to isolated the characters into several regions and i've resize
> all regions into a same size (60x12). Then I accumulate the pixel '1' by
> column and lets say that the data of pixel '1' is now in the size of 1x12.
> The same operations have been applied to all 10 images and I used all 10
> data (with size 1x12) to find the average and I used this average data as
> template image for character 'C'. Is that a proper way to do it? please
> advise because i'm still new in this area and that is why i'm starting
> with so-called simple technique. Thank you very much for any help.

It sounds to me like this is less a MATLAB question and more a general image
processing question -- if that's the case you'll probably receive better
responses on a newsgroup dedicated to image processing, like
sci.image.processing. If the feedback you receive there leads you to
need/want to make changes to your MATLAB code and you have questions on
that, feel free to come back with those questions.

--
Steve Lord
slord(a)mathworks.com
comp.soft-sys.matlab (CSSM) FAQ: http://matlabwiki.mathworks.com/MATLAB_FAQ
To contact Technical Support use the Contact Us link on
http://www.mathworks.com

From: Mije Jusoh on 20 Jul 2010 20:02
I'm sorry..guess I've posted this message to a wrong newsgroup. Thanks for your suggesstion.


"Steven_Lord" <slord(a)mathworks.com> wrote in message <i247lr$9lo$1(a)fred.mathworks.com>...
>
>
> "Mije Jusoh" <mspooh21(a)yahoo.com> wrote in message
> news:i23pq8$933$1(a)fred.mathworks.com...
> > Hi,
> >
> > I'm currently doing a project on car plate recognition which use template
> > matching and correlation value. I have read lots of other ppl works on
> > this matter and they have proven of achieving high accuracy rate. However,
> > my project is not like the others. I found out that the result is really
> > low where only few of the car plates have been recognized correctly. I
> > would like to know what are the possibilities that these wrong recognition
> > might occur? Was it because of I have prepared a not really good template
> > data? what is the correct way to prepare for the template data? Let say, I
> > have 60 images of car plates with character 'C' on it, and I just took 10
> > images of it (with variation of sizes). After segmentation process, I
> > managed to isolated the characters into several regions and i've resize
> > all regions into a same size (60x12). Then I accumulate the pixel '1' by
> > column and lets say that the data of pixel '1' is now in the size of 1x12.
> > The same operations have been applied to all 10 images and I used all 10
> > data (with size 1x12) to find the average and I used this average data as
> > template image for character 'C'. Is that a proper way to do it? please
> > advise because i'm still new in this area and that is why i'm starting
> > with so-called simple technique. Thank you very much for any help.
>
> It sounds to me like this is less a MATLAB question and more a general image
> processing question -- if that's the case you'll probably receive better
> responses on a newsgroup dedicated to image processing, like
> sci.image.processing. If the feedback you receive there leads you to
> need/want to make changes to your MATLAB code and you have questions on
> that, feel free to come back with those questions.
>
> --
> Steve Lord
> slord(a)mathworks.com
> comp.soft-sys.matlab (CSSM) FAQ: http://matlabwiki.mathworks.com/MATLAB_FAQ
> To contact Technical Support use the Contact Us link on
> http://www.mathworks.com
 | 
Pages: 1
Prev: Plotting , axes
Next: Inport structure