OCR character positioning problem

[andrake]

Hello Mr. Greg,

Are the images attached the "range of values" you were referring to? Thanks.

Andrei K.

[Greg_S-Video]

Hello Andrei

I was just looking for any values you had experimented with for parameters so that I could focus on areas that had not yet been explorerd.  Another thing that may make performing the OCR more successful is a higher resolution image.  Are you able to increase the resolution at which you capture the image?  I'd be interested to see if upsampling the image has any positive effect but I would need to experiment.  Please let me know if we can achieve a higher resolution image.  Thank you very much!

[andrake]

Hello Mr. Greg,

The Camera we used is Basler "piA2400-17gm" GigE. I think it's a 5 mega-Pixel Camera. I don't know if there's some setting to make the resolution better but as far as I'm concerned we already set the best focus for that small IC's. Attached is another image I was managed to capture. Thank you very much.

Sincerely,

Andrei K.

[Greg_S-Video]

Hello Andrei

Through some correspondence, it looks like the best thing to do in this circumstance would be to build up your OCR library.  You may have to identify characters on multiple images in order to give the software a better chance of identifying newer images.  There has only been one image trained in the OCR system so far correct?  Are you able to train for images into the system and then run additional tests?  Are you able to train more characters using more than some sample character?  The number "1" may need to be trained with multiple samples in order to account for statistical variance amoung your input images.  Please let me know if you are able to do additional training and testing.  Please let me know if there is anything else I can do or whether you would like me to look into anything for you.