Hi,
Running into situations fairly regularly where images to be recognized vary greatly in scale. For example, one page might be 5128 x 6681 pixels while the next is 3424 x 4454 pixels. This causes the locations of the elements to move in relation to my search areas. We use our in house imaging system to feed FlexiCapture so I'm at the mercy of the quality of the documents I receive. I'm wondering if anyone has any clever tricks to overcome the issue.
I've used "left/right of: pagerect.xcenter" and "above/below: pagerect.ycenter" fairly successfully but when the scale changes enough, elements start to be missed. On occasion I've had to make a second flexilayout to ensure everything is captured document to document but maybe there's a better way? Can Flexicapture automatically re-scale the images prior to recognition?
コメント
3件のコメント
Hello,
We cannot say much without seeing the used FlexiLayout, but we suppose that absolute dt coordinates should be avoided. Instead of it fractions of PageRect dimensions (like PageRect.Height/16 or PageRect.Width \* 2 / 5) could be used, or also some well-detectable static texts dimensions could be taken as reference (e.g. Below: PageRect.Bottom/6 + Header.stCompanyName.Height \ 2 ; )
Agreed, absolute search areas are out of the question. I'll give some of your suggestions a shot. I don't actually have a specific FlexiLayout giving me trouble at the moment, just looking to broaden my understanding on the subject in general. Thank you.
Update. We're developing a tool that adjusts scaling etc. prior to sending the docs to Abbyy for recognition. Where we pull our images from an in-house imaging system this solution seemed the best fit for our needs. That said, I run into this situation a lot and think it would be a great feature within Flexicapture...that is to say, some form of additional image pre-processing option to adjust scales, image widths etc.
サインインしてコメントを残してください。