Active LED "lens" shape

Post Reply
JeDi
Posts: 41
Joined: Thu Mar 10, 2011 3:59 am

Active LED "lens" shape

Post by JeDi »

Hi all,

We are using Tracking Tools to create a specific motion capturing and analysis setup for bicycles. Because of the reflective materials used on modern bikes, we quickly came to the conclusion that passive markers was not the way to go. So we are developing a setup using active IR LED markers.

Because of the specific setup, we only need six camera's (three on each side of the bike) and we can use bare LEDs on the body: the LEDs are all approximately facing the three camera's that can see them. This all works flawlessly.

But now we are trying to make the whole setup active, so we can use the cheaper SLIM cameras without IR light emitters, but also to ease the calibration process (with a passive system we have to get rid of most reflective materials in the capture volume). We need both the ground plane and the OptiWand to be active.

With bare LEDs put on the OptiWand, we didn't get much success: I had to point the LEDs towards the cameras in order for them to detect the point, which means that in our setup only three cameras could see the OptiWand points at once. We had the same problem with the LEDs we put on the ground plane.

So our next step was to use some sort of "lens" on the LED: a diffuse material distributing the IR light around. For now we used parts of the cylindrical sticks found in a glue gun, which seem to give good results. But, finally getting to my question, we are wondering how much of an influence the shape of the "lens" has on the tracking algorithms.


So in short, is it important to make the diffuse material around the LEDs as perfectly ball-shaped as possible? Does anyone know where to get more professional caps to put on the LEDs? Or any preferred material to make them?

Greetings,
Jeroen Dierckx
NaturalPoint - Mike
Posts: 1896
Joined: Tue Feb 01, 2011 8:41 am
Location: Corvallis, OR

Re: Active LED "lens" shape

Post by NaturalPoint - Mike »

Tracking Tools has a filtering option called a circle filter which can eliminate potential bad data based on the shape of the marker. The scale for the circle filter is 0 to 1 (not at all round to perfectly round). Generally this is set to .6, however during calibration it's set to .8. You can try reducing the circle filter to .6 (or lower if necessary) to try to get the cameras to recognize the LEDs as proper markers during calibration. What's likely happening is that the base of the LED of the part to which they're attached is blocking part of the light and making the marker appear out of round.
JeDi
Posts: 41
Joined: Thu Mar 10, 2011 3:59 am

Re: Active LED "lens" shape

Post by JeDi »

I'm not sure if I made myself clear in my post: the part we put onto the led is not round: it is a cylinder shape, cut from a stick from a glue gun. The material itself seems perfect for the job, but I was wondering if the cylindrical shape will be a problem for accuracy (so whether the centroid of the recognized blob will be correct).

I did a standard calibration without changing any settings, and the leds were recognized perfectly fine. We had a mean error of 1.5mm on the leds we put onto the OptiWand though, but we didn't place them perfectly centered on the leds.
NaturalPoint - Mike
Posts: 1896
Joined: Tue Feb 01, 2011 8:41 am
Location: Corvallis, OR

Re: Active LED "lens" shape

Post by NaturalPoint - Mike »

If the part is not spherical, it will have an effect on the position of the centroid. This can be compounded during caliration if the angles cause the markers to appear non-linear, as calibration requires a high level of precision. If you could shave the cylinders into spheres and place them on the IR LEDs, that would be ideal.
JeDi
Posts: 41
Joined: Thu Mar 10, 2011 3:59 am

Re: Active LED "lens" shape

Post by JeDi »

Allright, that was the answer I was looking for. I will find a way to shave the material as ball-shaped as possible.

Thanks
Post Reply