Change in form box size

Change in form box size

Postby JScotti » Mon Dec 04, 2006 1:09 pm

This is a general question and I've spoken to some techs who have told me this shouldn't be an issue. However, I can't seem to figure out another reason as to why this could be happening.

The form I'm trying to capture is an unfriendly one and by that I mean not scanner friendly. It can be viewed at: http://www.revenue.nh.gov/forms/cst/dp_135.pdf . The form we are using has many "box goupings". Each field has two boxes, one for the dollar amount and one for the cents amount. What I have done is grouped them together as a smart box and concatenated them with a ".". But, because many of the boxes are part of the others I've had to overlap them a bit (ie. 2a - 2e).

I have various zones created on the document and I have found that the recognition boxes are expanding on certain fields, none of which are set up identically. I figure this is due to Xtrata. My question is that because some of the fields are right on top of one another, they do overlap slightly.

Now my question. We've installed Xtrata (1.7), but we've still been having numerous problems with the lineup of the fields which is causing very poor recognition. I was wondering if the closeness of the box groupings, Xtrata cannot differenciate between one box and another. And if our boxes were seperated that the Xtrata could make a better recognition of the index fields.

Could this be the case? Or can Xtrata compensate for this situation on a form?

-Jenn
JScotti
Participant
 
Posts: 206
Joined: Mon Sep 25, 2006 4:00 pm

Postby russell@centuryc.com » Mon Dec 04, 2006 1:48 pm

I'm not sure why you need to overlap fields. I can see that you do need to put them right up to one another.

Something to keep in mind is that while Xtrata does very well with the normal distortions that scanning (and xeroxing) does to a form, it doesn't really cope well if someone has decided to reproduce the form in their own software. Are you really scanning the same form, or are these really different forms made to look alike to humans?

Since this is a tax form, it wouldn't surprise me at all but what some tax accountants have reproduced the form in their own software so their automation system.

Also, have you turned off ALL form registration so Ascent doesn't mess up what Xtrata has done? Don't use any registration point(s) and completely turn off any page registration.
russell@centuryc.com
Participant
 
Posts: 3374
Joined: Wed May 17, 2006 12:53 pm
Location: USA

Postby JScotti » Tue Dec 05, 2006 8:17 pm

I guess my thought was to get a better recognition, but now I am concerned that it is doing more harm than good. There are no registration points and page recognition I believe is set to zero.

Should I change the boxes so they do not overlap? Just so they meet. Could that make the difference?

-Jenn
JScotti
Participant
 
Posts: 206
Joined: Mon Sep 25, 2006 4:00 pm

Postby russell@centuryc.com » Tue Dec 05, 2006 9:01 pm

JScotti wrote:I guess my thought was to get a better recognition, but now I am concerned that it is doing more harm than good.


It's worth experimenting. I assume you drew the boxes larger to account for some registration slop. That's fine for some things, or if the writers tend to let their input go outside the box. But when Xtrata is doing it's magic, the registration should be perfect. If it's not, you need to find out why.

Double check the page registration setting on the Form Type Properties. It should be set to 0. By default it's 1%.

You might also check your scanner. I had one that had a bad belt. You could hear it clunk-clunk-clunking. At first the images looked ok, but on very careful examination, it could been seen that the document had a few crunched bands. Drove Xtrata batty. It couldn't even recognize the form much less register it.
russell@centuryc.com
Participant
 
Posts: 3374
Joined: Wed May 17, 2006 12:53 pm
Location: USA

Postby JScotti » Fri Dec 08, 2006 9:37 am

I drew the boxes larger because the index fields weren't lining up on the document. I was hoping to get a good auto-recognition.

What if there is no reason "why" it's not perfect? Is it possible that Xtrata just can't do what I need?

I think Xtrata can do the shift of the index fields that I need. Maybe it needs more sample pages or some type of spacing between the fields to tell one from another, maybe it's a little of both.

Definitly something to test out.

-Jenn
JScotti
Participant
 
Posts: 206
Joined: Mon Sep 25, 2006 4:00 pm

Postby russell@centuryc.com » Fri Dec 08, 2006 1:31 pm

As long as you are feeding it the exact same thing, and your scanners are healthy, Xtrata should be dead-on.

I've used Xtrata to process a survey. The survey was professionally printed in one run. The marks were little circles about 1/8" in diameter that were to be filled in. I drew one box per circle. Each box was just large enough to hold the circle. Xtrata dialed it in dead-on.

My biggest fear in your situation is that you actually have forms from multiple sources and that's causing the problem. That may seem identical until you put two together and hold them up to the light.

It's possible it doesn't like the forms layout. It's worth playing with.
russell@centuryc.com
Participant
 
Posts: 3374
Joined: Wed May 17, 2006 12:53 pm
Location: USA

Postby JScotti » Mon Dec 18, 2006 12:28 pm

There are forms from multiple sources. For testing though I printed some and the user printed some, and the copies were shifted, but Xtrata shifted to compensate. Which is fine. But I have others of the same form that taxpayers sent in where it doesn't shift. I think I might play around with the definitions and seeing if that would help.

-Jenn
JScotti
Participant
 
Posts: 206
Joined: Mon Sep 25, 2006 4:00 pm

Postby russell@centuryc.com » Mon Dec 18, 2006 12:37 pm

JScotti wrote:But I have others of the same form that taxpayers sent in where it doesn't shift.


Examine them very closely. Perhaps a company like TurboTax has created an equivalent form. The forms have to be the same form from the same source file. If it's being printed from other files that can't directly trace their parentage to the master, then you'll have a problem. Simply looking identical is not enough.
russell@centuryc.com
Participant
 
Posts: 3374
Joined: Wed May 17, 2006 12:53 pm
Location: USA

Postby stephan_mayer@kofax.com » Tue Dec 19, 2006 12:31 pm

Xtrata uses fuzzy pattern matching to register each zone individually, by looking at the surroundings of each zone.

Good image quality, especially for the first sample page, is very important.

To better analyse what is going on, you can use the Xtrata Classification & Registration Demo tools - available here (partners only).

As Russel already wrote, it is VERY IMPORTANT THAT YOU DISABLE ANY ZONE REGISTRATION IN ASCENT, otherwise this would overwrite the results from Xtrata.
- Set the "page registration" confidence (form type properties dialog) to 0%
- do not use any registration zones

Size and position of the zone you have set up is not really relevant, as Xtrata is automatically "locking" into the surroundings. If it does not work, either the offset is too much (believe we certified up to 4%) or there are no surroundings (e.g. color dropout).

Using the Registration demo you can easily check if registration works. Load your sample image to the left side, select the "not working" one on the right side, draw zone(s) on the left and then click the "register zones" icon in the toolbar.
Best Regards,<BR>
Stephan Mayer
Presales Manager EMEA
Kofax Image Products
stephan_mayer@kofax.com
Participant
 
Posts: 350
Joined: Wed Jan 28, 2004 3:40 am
Location: Germany


Return to Ascent Xtrata General Discussion

Who is online

Users browsing this forum: No registered users and 1 guest

cron