Hello,
some help with the following application is needed:
I have a litte robotics demonstration using a Mindstorms robot carring a USB Camera. The robot is controlled via bluetooth and the images are
captured via IMAQ USB - hardware setup works quite fine !
I need the robot to detect a target, which is a pattern very easy to recognize. A black X printed on a white sheet of paper (Lines are about 1cm thick,
X is about 12cm in diameter).
So I use a geometric pattern match (after some preprocessing) to recognize the X, this in general works, but I get only matches if I a allow a very
high "Obscured" value (about 60) and a match is only detected, if the image is in rectangular position to the camera (due to geometric distortion,
I think).
This is not useful for my application, because the robot should "see" the X, even if it looks on it with an angle of 45deg.
How could this be improved ?
I played around with some other pattern matching algos, but the geometric one seems the best for this
application.
Any help would be appreciated,
thanks in advance,
Markus