You are viewing limited content. For full access, please sign in.

Question

Question

check scanning cheque scanning

Quick Fields
Updated October 19, 2016
asked on October 18, 2016

Hello all,

Has anyone perfected cheque scanning? The font for the MICR enocded banking information is difficult for Quick Fields to read. The "1" reliably scans as "i" (which is good) but Quick Fields really gets tripped but by the other MICR font for letters. On the cheques they look like braille or mini barcodes. Quick Fields tries and gets 0s and 1s.

I've got a method but it's not reliable, working sometimes but not often enough for my liking.

-Ben

0 0

Replies

replied on October 19, 2016

Are you trying to read the Account Number? I had a process where I read the check number and then looked the check information up in a database to get all of the information on the check, however this was on checks generated by that company. This process obliviously would not work on checks coming from outside the company.   

0 0
View 3 previous replies
replied on October 19, 2016 • Show version history

Hi Derek,

In this image, there's a code after the 123, 456 and the 7. Laserfiche's OCR attempts to the recognise the symbols and gets it wrong. Worse, it's recognised as a number which messes up the account number extraction and lookup.

https://upload.wikimedia.org/wikipedia/commons/b/b9/CanadianChequeSample.png?1476890630192

 

 

0 0
replied on October 19, 2016

Have you tried a regular expression to try and get just the information that you need?

0 0
replied on October 19, 2016

I did. However those scpecial characters are recognised as numbers. Peppering in additional numbers and making the actual account number impossible to determine.

0 0
replied on October 19, 2016

This is more of a workaround, but you can get rid of some of those special characters with a Despeckle process. I tried with size 10 and isolated the 123 456 7. If you want to get rid of them all, you may have to use a couple Despeckle processes. Despeckle size 16 got rid of the remaining characters, but missed some that Despeckle size 10 got. So I used both and was left with just the numbers on the bottom line.

0 0
replied on October 19, 2016

Hoi Pieter,

On my sample scans (300dpi) I had to use 25px to get rid of the extra codes. It's looking better. 

Also, I've had to substitute "i" for "1", "O" for "11" and "?" for "7" so the question is, how repeatable is this process? I only have a limited number of cheques to test with. You're just being left with numbers, you say. 

I'm happy to trade images with you but not publicly.

-Ben

0 0
replied on October 19, 2016

I should mention that the numbers on one check reads perfectly and one doesn't I'll play around with other local enhancement options and see what comes out.

0 0
replied on October 19, 2016

I realize this probably isn't the answer you're looking for, but for anyone else who sees this, I'd like to reiterate what I said in this post, i.e. that Quick Fields with the built-in OCR engine isn't specifically built for reading MICR fonts, and thus there may be a limit to what kind of accuracy you can achieve without using other tools (likely via a custom process). 

That having been said, I would also love to know if anyone has a 'success story' to share, whether it was using Quick Fields by itself or using Quick Fields with a 3rd-party component. 

0 0
You are not allowed to follow up in this post.

Sign in to reply to this post.

Details
Related Posts
Loading ...