Forum

Welcome Guest 

Show/Hide Header

Welcome Guest, posting in this forum requires registration.
Pages: [1]
Author Topic: Visual calibration of polagraph geometry?
brandonagr
Newbie
Posts: 2
Permalink
Post Visual calibration of polagraph geometry?
on: October 31, 2012, 19:47
Quote

I just finished creating a polargraph that uses a Go program running on a raspberry pi http://code.google.com/p/gocupi/

My intention is to use it to draw on whiteboards in different locations, so ease of setup and calibration is a concern(which is why I went with just a spool of thread and no counterweights). Currently I have to manually measure the length of each string and save that in a config xml before running the program, but this seems very error prone and time consuming.

So I was wondering if anyone has looked into having the polargraph draw a known pattern, and then measuring the distortion of that pattern to solve for what the initial string lengths actually were to begin with? It would start out with some rough estimates for what I think the geometry is and draw the shape, then I would have to calculate what the actual geometry of the motors / gondola is based on any distortion in that drawing. Does that sound like a plausible approach?

The other auto calibration approach I have seen is to have some kind of sensor that detects when the gondola is pulled all the way to the motor, but that would add more mechanical complexity and wires to the setup, so if I could precisely calculate the geometry just by measuring a simple pattern that seems like an easier approach.
sandy
Administrator
Posts: 1317
Permalink
sandy
Post Re: Visual calibration of polagraph geometry?
on: November 1, 2012, 23:02
Quote

Hi Brandon, I meant to send you a message about your project - beautifully even and regular results.

I did think about having a kind of printable calibration test patch that would somehow tell the machine a few things about where it is, but the stumbling block was really how to get that information into the machine. Video is the logical way to do it, but that's beyond arduino. It could pipe a feed into the processing app and calculate the machine size that way but I wanted something standalone. Actually, with any kind of cv approach, it'd be easier just to spot the positions of the motors and the gondola in the first place.

I mean, this is all ignoring the fact that I have no really clue about geometry or the kinds of maths that would be required to make use of those measurements anyway. That's really why I've tended to focus on the mechanical end-stops idea. More wires, more hardware complexity, but also more tested solution, and I can immediately see how the calibration would work in my head.

Kind of the closest I got to thinking of a non endstoppy approach was a boundary kind of thing, where a reflectance sensor on the gondola would be triggered by a dark just drawn around the surface, or some special fiducials or something. So calibration would involve the machine exploring the area until it detected all of the edges and corners, and then somehow do mapped its own coordinates system onto the cartesian space it has just discovered.

Interested in talking about this though.

cheers!
sn
brandonagr
Newbie
Posts: 2
Permalink
Post Re: Visual calibration of polagraph geometry?
on: November 2, 2012, 19:16
Quote

Yeah its getting the feedback in an automated way that is difficult, the boundary detection on the gondola where it could tell if its at an edge is an interesting idea. Since I am planning to eventually use my phone to remotely send commands to start it running, perhaps I could use the phone's camera to take a picture of the drawn out calibration pattern and then send that to the computer over wifi to analyze the pattern. There would probably have to be a separate paper printed calibration target in the picture so that the pose of the camera could be determined before then analyzing the calibration target drawn by the polargraph.

For now I added a mode where I can type in the current X Y location of the pen, and then it calculates what the current string lengths are, that way I can just measure horizontal / vertical position which might be slightly easier than having to measure the string lengths
sandy
Administrator
Posts: 1317
Permalink
sandy
Post Re: Visual calibration of polagraph geometry?
on: November 4, 2012, 19:02
Quote

The phone thing would be a great idea - IR reflective dots on the motors and gondola, and draw a square of a fixed size. So the dots would indicate the shape and current gondola position, and the pattern would indicate the size of the setup. It's beyond me, but it does seem like it could be a very elegant way of calibrating a machine, if it could be done accurately enough. To be honest, I've tried to be accurate down to the millimetre, but I bet I could get away with a lot less precision in practice, though it depends on the results I suppose. The stuff you're drawing seems very square and true.

sn
Pages: [1]
Mingle Forum by cartpauj
Version: 1.0.34 ; Page loaded in: 0.049 seconds.