Is anyone using Advance Capture for many OMR fields/values?

[#1]

Good day,

I have been struggling with getting Optical Mark Recognition (OMR) to work reliably.  Support has been their usual helpful selves, but we still have not made sufficient progress to make OMR to be a workable solution.  I am interested if anyone has forms that are strictly OMR and if they have been able to get them to work consistently.  Using Kofax in the distance past with a similar form (that is, only OMR on it taking up the full 8.5x11 sheet) the error rate was much less than 1%.  The best results I have had since August is a 20% error rate - far from acceptable.

OnBase info:  OnBase V13.0.1 core running V13.0.2 thick client.  Advanced capture is running as a service (and therefor as V13.0.2) using the Hyland OCR Engine 18.62

Form info: Form Size: 8.5x11 inches, simplex.  Fields:  17 OMR groups with six to 13 check boxes.  The form is always printed from the same machine using the same file - it is not photocopied.

Configuration: Each group is configured to have a minimum of one and a maximum of one positive result.  Mark frame detection is set to Framed.  Support 'filled in error' detection is unchecked.  Sensitivity is at the lowest, as users are now coloring in each check box rather than just checking or marking an 'x' inside the box.  

Scanner: Fujitsu fi6670 running paperstream with deskew and despeckle turned on

problems: Advanced Capture seems to take the beginning of the form to be the upper left corner of the paper rather than the form identification fields/values. if the form is not inserted into the scanner in exactly the same position, minor changes in the scanning position/process essentially mean that most of the groups become either empty or most values are identified as marked.  Using the configuration utility we can see the zones moving around the page, even though the document has been identified by the identification zones.

Another problem is that when I test the OMR group, it finds a single value.  But when the same document is processed a group that was shown in configuration/test to identify a single positive mark will find two positive marks.  This has been reproduce on the Hyland side, though no explanation/solution has been found.

After thousands of real world documents and multiple times a week of tweaking the configuration - including many sessions with Hyland Support with multiple people, we are finding that OMR may not be able to handle what I thought was a pretty simple form.   Has anyone had success with forms that are designed to be only used for OMR?

thanks!

[Chris_Bennett1 · Accepted solution]

Hi Dave,

From the description of your issue it sounds like the document’s shifting is causing the zones to be slightly off, which in turns is causing misread values. Fortunately OnBase does have the ability to account for slight variations in the document due to the unpredictability of scanning.   The best way to account for document’s shifting is to apply multiple Form Identification Zones. Place a Form Identification Zone in opposite corners of the document. (One in the top left and one in the bottom right, or vice versa) This will allow the template to properly scale itself. Also each form identification zone will need to use the option ‘Use registration point.’ The template will look for each of these registration points then scale itself to match each point. This is why using two form identification zone is better than only using one. For more information on template scaling you can look at the OnBase 14 Advanced Capture MRG page 225 or the OnBase 13 Advanced Capture MRG page 204, under the section title Configuring Multiple Form Identification or Page Registration Zones for Scaling.

[#3]

Hi Chris,

I see that this was marked as the answer. Despite many hours, OMR is extremely picky. Ideally forms would always be printed from the same printer, that always has just been serviced and are scanned with the paper guides perfectly adjusted to prevent skewing in the scanner. Our real world experience is that these three things rarely come together and even with three to six recognition forms in different sections/corners of the document, the system has a difficult time correctly recognizing marks. And not just marks. We have experimented with coloring in the boxes rather than just checking the box resulting in a very large black spot, but even then it has problems recognizing the correct mark in a group.

Due to this difficulty and zero responses from other OnBase customers using OMR we are no longer pursing OMR as a viable import strategy. If (advanced) capture gets a better engine/upgraded logic we may revisit this option.

[#4]

Hi Chris,

Thanks for your thoughts.  We have had multiple form identification zones (two to four), all with registration points ,in various locations on the document, though there were never any available in opposite corners.  We have recently added an additional identification zone in an opposite corner and are waiting to see what the effect will be in the long term.  Initial testing is good, but so was initial testing when we had two and three form identification zones at various locations on the document.

With the Hyland's help, I have been working with Multiple Form Identification and Page Registration zones for scaling - that is what got us to 'only' a 20% error rate.  Hopefully the addition another form identification zone in the lower right corner, combined with the existing one in the upper left corner (as well as the three others that are available throughout the document) there will be much better results than we are getting now.  

Thanks for your time, much appreciated!