forums.naturalpoint.com

Hi,
I intend to develop an application using Optitrack cameras.
In the application, I try to test if a rigged body was moved -tiny movments.

 X, Y (& Z)
 Resolution � 1/4 mm � 40 mm
 The rigged body length is about 1 meter

Is it possible to set up Optictrack camera\s at any position it seems the best, to answer those requirements?
Testing my requirement with Optictrack Flex 3 legacy cameras, did not achieve the desire output.
The major problem except of the accuracy was the �crawling� behavior of the position output when the rigged body was static (hundredth, tenth and event got to units. The test was done at a dark room).

My questions are:
Accuracy
1. What are the performance accuracy of the Flex: 100 \ 120?
2. Is there any way to achieve better accuracy:
a. Using multi cameras?
b. Position?
c. Lens?
d. Lighting?
e. Markers? Active?

Crawling
3. Would the crawling behavior will happen with the new
cameras.
4. If crawling does occur � how can I reduce it or detect it in my application?
a. Any of your software can help me with this problem (Point cloud, Rigged body)?
b. Any other smoothing algorithm?

Regards,
Mat

It sounds like you may be seeing the impact of the smoothing algorithm on the data. Have you tried using the raw data instead of passing it to the smoothing function first? The trade off is that some jitter in the data will be present, but you could choose to discard below a certain decimal place to work around that.

In overall terms, the Vector tracking algorithms are tuned to work best when used with TrackIR cameras for 3D view control, although they will still provide reasonable estimates of head position for OptiTrack cameras. We do not have hard numbers for the performance, and there are not parameters which can be adjusted to improve it. If better resolution and accuracy are needed, we would recommend looking in to the Point Cloud toolkit or the upcoming Rigid Body toolkit. These provide more robust tracking and a wider range of motion.