I want to setup a simple OCR for LCD displays using phone camera. Any idea where to start from?
I want to setup a simple OCR for LCD displays using phone camera. Any idea where to start from?
Hi Cassioli,
I was also planning to start same project to recognize OCR, for visually challenged. If you have no problem i can work upon this.
IDEAS is all they need but still they think only Genius can give them that.
And i think i got what you are thinking of doing first take the image, convert it into black and white, the LED have intensity so they come in white part and taking the pattern you recognise the Character? Am i Right?
IDEAS is all they need but still they think only Genius can give them that.
thanks, it would be cool.
Actually I'm not working to such a complex project, I just need to recognize LCD figures to store them like text for further processing (for statistic purposes).
But it looks like just examining 7 areas of the screen for dark pixels is not enough to determine which segments are on, due to different lightning conditions over various areas of the image....![]()
Your Idea is really a start line for such a project.
I think, it would be better to look for 7 bright parts of image. I am sure that using White LED or somehing like Blue LED will have high intensity as when image taken (even when converted to black and white), and ya its 8 parts not seven, because 7 segment display also contains a Dot which is 8 part.But it looks like just examining 7 areas of the screen for dark pixels is not enough to determine which segments are on, due to different lightning conditions over various areas of the image....
IDEAS is all they need but still they think only Genius can give them that.
I completely forgot the dot...
Anyway I can't look for bright areas, as I am trying to decode a real LCD display, nota a LED LCD display: it's an indoor/outdoot thermometer.
BTW, is there a method in python to automatically convert an image to b/w, rather than examining it pixel by pixel?![]()
Ya changing an image to Gray Scale can be done by help of Open GL functions
http://www.gamedev.net/community/for...opic_id=456541
and i think PyS60 can bind to OpenGL.
IDEAS is all they need but still they think only Genius can give them that.
Found a better method to save in b/n:
img.save("nome",None,None,100,1)
I attach the source I'm currently working on.
pys60 1.4.5 and 2.0.0, pygame, PyS60 CE on E90 and 5800 !
Find my pys60 extension modules on cyke64.googlepages.com
Thanks for the big news!
In the meantime I put my OCR project an googlecode, for anybody who wants to help...
http://code.google.com/p/nokdat/source/list
I'm working on 2 versions:
- old one just examine 7 areas of screen, by summing up all pixels values and comparing sum to a user-defined threshold: if greater, than white, i.e. segment off.
This version appears not to be working on different light conditions, so I'm now trying to examine surroundings of each segments: if surroundings are brighter than segment itself, segment is on, else off. This version also saves a (long) log file to examine the processing.
Any help is appreciated:how could I muse new libraries to improve my project?
My google code page:
http://code.google.com/p/nokdat/source/list
pys60 1.4.5 and 2.0.0, pygame, PyS60 CE on E90 and 5800 !
Find my pys60 extension modules on cyke64.googlepages.com