RoboRealm Forum

No Title

Tue, 30 Nov 1999 00:11:00 EST

Arnie,

Can you include the original images in this post? Its hard to know what's going on using the screen shots. Including the original unedited image is the best way to determine the issue.

Also, I've noticed that you are not segmenting the letters prior to use. Normally one uses the Adaptive Threshold or perhaps the Auto Threshold module before the OCR module to ensure that it is processing a black and white (binary) image. That might be part of the issues you are seeing where the letters are being interpreted in smaller pixel sized objects since the image isn't black and white.

STeven.

No Title

Tue, 30 Nov 1999 00:11:00 EST

Hi STeven,

Thanks for your reply. I've tried your suggestion of increasing the Word Separation up to 1000 which did help but I'm still having issues the number '4' in particular in that it seems to take priority (moves to the front of the captured value) as seen in my examples I've attached. One of them is a 'good' value for comparison.

Any ideas?

Thanks,
Arnie...

No Title

Tue, 30 Nov 1999 00:11:00 EST

Arnie,

There is an issue of how the numbers are being reported that we are looking into. In the meantime, you can avoid that problem by increasing the 'Word Separation' value in the OCR module to about 200% which will merge all characters into a single word in the right order.

STeven.

OCR Result - Number Orientation

Tue, 30 Nov 1999 00:11:00 EST

Hi STeven,

I am successfully using the OCR module to capture anywhere from 5 to 9 numbers (quantity of numbers changes randomly) in a row. I expected the OCR result to represent the same sequence orientation captured on the screen but what I get is a random sequence which seems to be affected more by particular numbers over others. For instance, 101.72 appears as 01172 where 101.62 appears as 62011. I've included a sample program and some pics.

Is there a way to capture these numbers in their correct orientation every time, independent of their value?

Thanks,
Arnie...