Page 2 of 2

Re: IR Tracking Prototype (with proof/image!)

Posted: Fri Nov 23, 2012 6:41 pm
by WiredEarp
Re using color, as mentioned here:

cadcoke5 wrote:
Another way is for one of LED's to be a distinct color from the rest.
That's going to be hard to do with IR I guess, isn't it basically B&W for the camera ?

Why not just use two cameras? One responsible for tracking the IR markers, and the other one can be used every X frames to detect the colors and assign individuality to the groups of markers.

Re: IR Tracking Prototype (with proof/image!)

Posted: Sat Nov 24, 2012 9:19 pm
by Nick3DvB
Filters are a good idea; maybe even simple anaglyph filters on the cameras normal red/blue LEDs would work? I don’t know if you can get very narrow band IR LEDs / filters but I wonder if you could use circular polarization filters from cinema glasses in front of the LEDs / Camers? They might cut the range down a lot though.

I’ve not had chance to test the new build yet, I knackered one of my PSeye cams trying to do the IR mod, then got side-tracked with a video player idea, but as soon as I can get hold of another camera I’ll give it a shot. Keep up the good work Memebox! 8-)

Re: IR Tracking Prototype (with proof/image!)

Posted: Sun Nov 25, 2012 4:38 am
by MemeBox
I've got my arduino and hooked up an led. I'm currently outputing 2 different sine waves concurrently, the second wave changes frequency every 3 seconds. I have a ps3 eye tracking the led and recording the brightness values of the tracked point over time. I've got a graph of the fourier transform of the recorded points and I can see the peaks of the graph move as the arduino changes the sine waves in the led. Awesome. So all I have to do now is find the similarity of the current frequency vector to a list of different vectors for each led and I can tell one led from another. I'm hoping this will also help with tracking, since if bright points in the image coming from a light source which is not an led are registering as something to track I can discount them by observing that they do not have the correct spectrum in the frequency domain, that they are not composed of the correct frequency components :) I'm not sure how much cpu power this will require, hopefully it will be ok, but if not I'm considering offloading it to a gpu with custom shaders...

Re: IR Tracking Prototype (with proof/image!)

Posted: Sun Nov 25, 2012 4:48 am
by MemeBox
Oh I'm also considering moving to the visible spectrum, this way I could use, as suggested, rgb leds. This would give me 3 channels over which to broadcast the identity of the led. This would allow you to identify the led quicker and probably track it more easily. I'm tinking that if someone is prepared to put up 3+ cameras in a space they won't be too fussed about flashing lights. I'm also thinking earnestly about vr laser tag, where it doesn't matter a jot whether you can see the led light or not. I'm thinking that I could use a raspberry pi and ps3 eye for each camera, do the image processing on the pi and then forward only the tracked points to a central server via wifi. Then the central server could forward the appropriate details for each player over wifi to a raspberry pi for each player which would render everything for each players rift...