Figuring out Angle from X and Y values?

If it only moves in X and Y, there would be no angle to calculate, only an X and Y offset.

Maybe I am not understanding your question.

So you are talking about drawing a line from one X,Y position and calculating the angle of that line? Maybe a picture would help me understand.

I am still not following fully. A diagram would really help. If you have multiple cameras with overlapping views, I would recommend you calibrate the cameras together first so they are dialed into the same grid.

From a math standpoint, the angle is arctan((y2-y1)/(x2-x1)). I think that to implement that on the robot, you'll need either the math option or to use Karel.

Quote from TitusLepic

From a math standpoint, the angle is arctan((y2-y1)/(x2-x1)). I think that to implement that on the robot, you'll need either the math option or to use Karel.

I've had some success using Taylor expansion to get around the math limitations on base fanucs. You have to scale the function though, to account for the fact that a real in the registers is cut off after 6 decimal places.

Einstein would be interesting with this challenge

They way we (Einstein and I) see it , it's that the cameras are in parallel worlds. The coordinates are totally independent from another
I don't believe the robot knows exactly where the corners are related to each other, it know for example, one is at -1 and the other +1, but + or - from what ?

I think you have to put an object where you think, it's at zero degree (this is the big deal) .Then take the pictures and then yes, whatever result you get, it is your zero and you can follow any line equation to find the angle such as y-y1=m(x-x1)+n

I can tell you right now is not going to be perfect, but you can adjust it the best you can

Quote

I've had some success using Taylor expansion...

Brilliant. This is going to help me a lot, thanks.

It does not cost mutch too add the math function option, witch gives you trig, exp, etc...

Sent from my VS985 4G using Tapatalk

If (x₁₀,y₁₀) is your zero point of camera 1, (x₂₀,y₂₀) is your zero point of camera 2, (x₁₁,y₁₁) is your recorded point of camera 1, and (x₂₁,y₂₁) is your recorded point of camera 2, the difference in angle will be arctan((y₂₀-y₁₀)/(x₂₀-x₁₀) - arctan((y₂₁-y₁₁)/(x₂₁-x₁₁)

Quote from TitusLepic

If (x₁₀,y₁₀) is your zero point of camera 1, (x₂₀,y₂₀) is your zero point of camera 2, (x₁₁,y₁₁) is your recorded point of camera 1, and (x₂₁,y₂₁) is your recorded point of camera 2, the difference in angle will be arctan((y₂₀-y₁₀)/(x₂₀-x₁₀) - arctan((y₂₁-y₁₁)/(x₂₁-x₁₁)

Don't you need to know the distance from one camera to the other?

I'm not too sure how vision works and is set up, since I've never used it. As long as all points are relative to the same origin, this will work. If each camera's points are relative to that camera and the camera doesn't know where it is relative to the origin, then yes, you'd have a add each camera's location to its offset in order to get coordinates to work from.

Maybe I'm missing the point, but isn't this one of the functions of a multiview/camera setup in iRVision?

Are your parts always a fixed size, as far as the distance corner to corner?

I tried to think hard an figure out some ratio of how the x direction changes relative to the y direction. I wanted to use this to take a percentage of the known angle, but it comes down to approximating the trig functions. That Taylor Expansion sum with the squared and cubes of the position appears to be the way to go.