Problem with Read/Verify Text in VBAI

[thangnguyen]

I've a big problem in using Read/Verify Text step in VBAI: VBAI always have a mistake between C and c; O, o and 0; B and 8; S and s; V and v; !, ., l, i and 1.

Could you help me to find a solution for this problem?

I'm using Basler camera to capture text of image. 

[muks]

See whether this will help

[thangnguyen]

I've tried with 2 methods but it's NOK because there's some characters can be recognized exactly and I've to use characters and numbers at the same time.

[muks]

Can you post a few sample images. May be I can try enhancing it a bit.

[thangnguyen]

This is 2 images about my text. Could you please help me to find a solution to improve it? 

[thangnguyen]

This is my OCR training file.

[BruceAmmons]

Check if you can limit the size and ratio values in OCR.  This might force it to be a more exact match in size as well as shape.

Bruce

[thangnguyen]

I've tried all parameters to improve quality of OCR but always have mistakes between i, l, ! or C,c or V, v or S, s.

Do you have any more solution for this? 

[A_Person]

Can you post the original image without the ROI and OCR text?  I would like to give it a try and make sure the overlays are not affecting anything. Also, some indication of the parameters you are using might help us make some recommendations.

Unfortunately, the OCR algorithm can have some trouble in the cases where you are having trouble.  You can clean up the image before performing OCR on it, performing some of the changes recommended by other posters, and of course, training it with more images.  This last option might help it differentiate the sizes better.

Message Edited by A Person on 10-21-2009 10:13 AM

[BruceAmmons]

Setting the aspect ratio close to 100% may improve things somewhat.  This will make sure the size and height/width ratio are both close to the original character.

Bruce