Machine Vision

Showing results for 
Search instead for 
Did you mean: 

Detect dark 'needle' on light background (360°)

Go to solution
I'm not sure but you can try overlaying Gauge Indicator on to image and make the needle invisble.
Message 21 of 27

Haha, it's funny that you mention it, because I (as an beginner) tried that and it worked, BUT it is far too inaccurate, is there an different solution?

0 Kudos
Message 22 of 27

never mind, It is not important anymore:)


0 Kudos
Message 23 of 27

Now it becomes interesting, the deviation is at some points too much, as well negative as positive. That's because of the resolution. Is it possible to increase the resolution, p.e., make from every pixel four??. (pixel overlap?)

0 Kudos
Message 24 of 27

It isn't the resolution.  It is more a perspective thing.  For accurate measurements, you need to be looking directly at the needle from a perpendicular position.  The needle has to be flat against the surface, and it needs to have the same shape in any position.  The lighting has to be good so there are no shadows.  I would not expect to get fantastic measurements from your setup, so it may be the best you can get.



Bruce Ammons
Ammons Engineering
Message 25 of 27
He gives at 8 points the correct value, at 12 oclock, 3, 6 and 9 amd between them, so the circle isnt really a circle, and thats because of the resolution, so it aint gonna be exact, but the higher the resolution the less the deviation. If I cant increase the resolution, I will try multiply the array with a handmade array as a filter. (Basicly I thougt that every pixel is a constant N degrees, but at some points (8) there are more pixels in the array), so the deviation is because of the calculation based on a constant index/degree
0 Kudos
Message 26 of 27

I did it!, I created an array (sine waveform) and added it to the other array to compensate the deviation, and it worked perfectly!:) The max deviation now is only 1 degree, thanks everyone for the responses

0 Kudos
Message 27 of 27