Get rX rY position of an object

[Rickvb]

Hello,

Currently I am doing a internship project. But I keep getting stuck.

What I am doing:

This is a Stewart-platform. The platform can be moved in 6DOF. The actuators are muscle cylinder without bearings.
I am trying to get the xyz + rotation angles(rX rY rZ) of the platform. I do this by using 1 camera (Ueye) which looks at the bottom of the platform, on the bottom there is a pattern image:

What I do is: I throw a threshold over the camera image, and use the particle analysis for the 5 point, to get there centre of mass and area.
The idea of getting the rX and rY of the platform was looking at the pattern, and measure the deformation which takes place.

what it looks like in 0position:

What it looks like in pure rY movement:

What it looks like in pure rX movement:

What is my problem:
I try to measure the deformation of the rX and rY which will look like this (extravagant drawing):

Every measure I made (length and area) in a pure rX position (see picture 5) will also change the measure of rY position.
So in my measures it looks like rX and rY are changing, but this is not the case. Only rX should change.

Now has somebody any idea of making my measurements more accurate, so I can say that this is an rX movement and this is an rY movement??



I hope you understand my post. It is really difficult to explain things in a foreign language.

Sincerely,
Rick

[Klemen]

Hello,

does the platform really move in "pure" rX (or rY) motion? What is the size order of the motion (quantiatively speaking (cm, mm...?)

Have you tried to correct for the perspective projection/distortion and also lens distortion?

Also, I do not see how you will be able to extract the rotation angles? For example in the rZ direction, you will probably use the thresholded area size (or not?), so your rZ motion will be somehow proportional to the object(s) scale? What about rotation? Can you have multiple combined transformations at once (for example 2 rotations + 2 translations)?

Best regards,

K

[Rickvb]

Thankyou for your intrests Klemen!

No it maybe not move in "pure" rX motion. But the measurements I made looked like rY also moved almost linear according to rX in a rX movement. This is surely not the case.
rZ will be extracted by using all the 5 points centre of mass positions. But rZ is not a really important case right now. I really want the rX and rY. The rZ is really small, so it will not interfere much in  rX and rY directions.

I already tried a lens correction for distortion, this didn't madee any difference. The lens distortion just made the measurements more precise(more deformation).


best regards,
Rick

EDIT: The movement is alot (more then 10mm). I have not measured this.

[Rickvb · Accepted solution]

I got my answer!

What I do now is:

I calculate these 4 surfaces with all the 5 centre of mass points.
This way i get the biggest deformation delta values. It is already better then my gyroscope on my phone!
Tommorow I do it with a more precise measuring device, but this looks promissing.

Thank you for your intrests, unfortunate we couldn't come to any solution.

Rick.