forums.naturalpoint.com

Hi all,
I am currently able to stream optitrack pose estimate of a rigid body using vrpn onto a computer running linux. I want to find the covariance involved in the estimation done by optitrack. The main problem I am facing is that the high frequency noise involved in the estimate seems to be higher when the rigid body is moving and even higher when some of the markers are blocked. The vrpn data just blindly gives an approximate pose estimate and doesn't provide the variance measurements. There is a kalman filter in the tracking tools gui but that doesn't provide any information outside the api.

Can anyone suggest any method for evaluating variance of the optitrack data using vrpn or any other means of knowing the reliability of the measurement I get through vrpn streaming. I dont mind using other streaming methods as long as they are compatible with linux. (I need this data for using in an extended kalman filter with imu measurements and other sensor data).

Hi ggarime1,

As far as real-time stream goes, the only place you're going to find any information about the quality of the pose for a given frame is in the NatNet streaming output. In addition to the rigid body pose, marker positions, and other rigid body solution specifics there is a 'mean error' metric that is streamed out. This is meant to be a measure of how good the solution is for a given frame. It's the average error of marker distances between the rigid body definition and the tracked rigid body on a given frame.

In our latest NatNet SDK you might want to check out the Unity sample. It will parse the NatNet stream and relay it over UDP. It wouldn't take much to port the slipstream.cs to whatever language you're using to receive data in Linux. Just an idea.

Also, you might want to check out our latest release of Motive (v1.5). It's running a new rigid body solving engine and I think you'll find it to be highly improved over the Tracking Tools solver.

Thanks,
Doug