Hi,
I’ve created a quick console app which takes a scanned pdf document of a questionnaire (dynamically created based on entries in a database), and an AMR file (again dynamically generated on the fly). When the scanned document is perfectly aligned it pulls the results I’m expecting, however, if the scanned document is a bit to the left or right of the AMR elements the correct values aren’t picked up.
I understand we’ve now got autoskew methods, however this doesn’t seem to make a difference to the outcome, and just fixes documents which need rotating.
I have the opposite problem in that the scanned image requires nudging up/down, left or right.
I’ve been looking at a way to incorporate some scanning marks like a black square in the corners of the document so when processing the image I can automatically rotate and adjust the position of the document so it matches the AMR file elements perfectly, but I’m struggling a little.
Any ideas / help would be welcome
Thanks
Lee
Hi Lee,
Thank you for your inquiry.
This is to update you that we need problematic image as sample and the sample code. With the help of these we will try to reproduce the issue at our end and analyse it.
Hi,
As requested here's the three example files (I've had to zip up the .AMR file for some reason you can't upload that file type here).
I've also had to edit out specific parts of the document due to sensitivity, but the elements, page size etc all remain intact.
You'll see by loading to AMR file into the OMR Template editor and adding the example dynamically generated page the overlay is fine (probably could do with some slight tweaking but it is dynamically created so a good job so far).
But, if you then load the scanned in image the image is skewed still by a few degrees and you can see a black edge on the right which means the scanner has taken that as part of the page, there seems to be additional spacing at the top too.
All this means the AMR overlay is miss-aligned so won't pickup the results. I can't ask the company to replace dozens of scanner as that's not an option (lots of other software / scanning solutions taking place) - and they're good machines.
That leaves me to try and compensate, so I was looking at how other OMR solutions get past this, and they tend to us either black markings down the edge, or crop marks, and the OMR engine picks these up and can auto straight, crop etc...
That would be a killer feature, but something that looks to be lacking here. So I'm thinking of adding something I can "recognise" in OCR / Image Processing and manually perform the alterations, but if you've any better suggestions or alternative methods that would be great as you'd expect something like this to be inbuilt as it's bound to happen to the majority of customers.
It needs to be a rock solid solution in order for me to deploy this, and I've got issues with text recognition too, but I'll work around that and just concentrate on this issue for the time being.
Thanks
Lee
Hi Lee,
We are investigating the issue. We will soon update you about our findings.
Hi Lee,
This is to update you that we investigated the issue at our end. We are able to notice the said behavior. The issue has been logged into our issue tracking system with ID OCRNET-2798. Our product team will further look into it. We will update you about the updates via this forum thread.
The issues you have found earlier (filed as ) have been fixed in this Aspose.Words for JasperReports 18.3 update.