sikuli-driver team mailing list archive

[Chris Kiick <question228095@xxxxxxxxxxxxxxxxxxxxx>]

New question #228095 on Sikuli:
https://answers.launchpad.net/sikuli/+question/228095

Here's the problem:  region.text() is not working for me.  This is a known problem with the OCR engine, I know.  It's probably due to the small font size that I have to deal with, the text color and the noisy background.  Whatever the reason, text() either gives nothing at all or semi-random characters, and it's not at all consistent.

However, I can do something based on the color of the text (white, yellow, red).   But how do I get sikuli to recognize/know the color of the text?  I can't use exact matches (even if that worked) because the text changes.

I thought of two ways to do this but don't know how to implement them.

One is to scan the entire region for a particular color.  Except how do I capture the font color in question?  What if it's not a single, simple color but has shading?  Also, there would have to be some kind of threshold count so that random background pixels would't give a false positive.  I'm sure I'm not the first person to have this problem, so are there any scripts out there that do this?

The other thing is to get a snapshot of each character (it's a number, so only 10 characters) and try to do matches on each digit, in each position, for each color.  For a 5 digit number that's 5 * 10 * 3 = 1500 matches.  Doesn't sound feasible. Plus all the coding to handle shorter numbers.  Unless there's a trick I'm missing?

Any suggestions on how to handle this?


-- 
You received this question notification because you are a member of
Sikuli Drivers, which is an answer contact for Sikuli.