Question #256236 on Sikuli changed: https://answers.launchpad.net/sikuli/+question/256236
RaiMan proposed the following answer: OpenCV features used in Sikuli: currently only matchTemplate is used (+ some Mat ops to adjust or detect image contents). This wil stay the same of course in SikuliX 1.2, but will be enhanced in various aspects like ignoring transparent areas (or even colored spaces), allow masking, use edge detection for special purposes like buttons … … and optionally try to evaluate the remaining differences in situations with a score less than some max (standard 0.9999 which is taken as exact) The other point is to support optional automatic optimization of screenshots in the above mentioned sense. Again: the more mainly even background the reference image (the image to be searched) has towards the edges, the lower the resulting match score will be (variation about 0.05). example: the arrow in the first sample will match with 1.0 if it has less than 20 % black surrounding pixels. The more black pixels around, the lower the match score (it might drop below 0.95 in some situations). This is simply due to the statistical functions that are internally used in OpenCV's matchTemplate. So currently the strategy should be: 1. have optimal shots, that get match scores above 0.98 2. restrict the region to the area, where the match should be 3. adjust the similarity used for searching to a higher value or even 0.99/exact() by using Patterns (not default similarity) -- You received this question notification because you are a member of Sikuli Drivers, which is an answer contact for Sikuli. _______________________________________________ Mailing list: https://launchpad.net/~sikuli-driver Post to : sikuli-driver@lists.launchpad.net Unsubscribe : https://launchpad.net/~sikuli-driver More help : https://help.launchpad.net/ListHelp
