problem with streamed rigid body data from Area

Post Reply
rranon
Posts: 5
Joined: Wed Sep 28, 2011 4:45 am

problem with streamed rigid body data from Area

Post by rranon »

Hello,

using the latest Arena ( 1.7.3 ) with 8 V100:R2 cameras, I stream rigid body data to my application, using NatNet 2.2.

The issue I am encountering is that the geometry to which the streamed data are applied oscillates even when the tracked rigid body is still. A colleague of mine, using the same configuration except for Tracking tools instead of Arena, is not experiencing any issue.

Is there any difference in the data streamed from Arena and Tracking Tools? E.g. Tracking tools is using meters while Area is using millimiters ...

What kind of precision can I expect from my setup? E.g. is it normal to observe an oscillation in the range of millimiters for a still rigid body?

Thanks for any help,

Roberto
DonBurroni
Posts: 127
Joined: Thu May 27, 2010 2:26 am
Location: Manchester, UK

Re: problem with streamed rigid body data from Area

Post by DonBurroni »

I have arena 1.7.2 and i get the same oscillation, but less in the volumes centre, even with the VCS plugin too, I have to filter the data. I think tracking tools is better software, with more features for calibration.

you can try to reduce the residual in he tracking features, also maybe even try, point cloud V1 too.
NaturalPoint - Mike
Posts: 1896
Joined: Tue Feb 01, 2011 8:41 am
Location: Corvallis, OR

Re: problem with streamed rigid body data from Area

Post by NaturalPoint - Mike »

When you refer to oscilation, on what scale are you getting it? How many millimeters are they moving? Is the oscilation worse in certain areas of the volume?
rranon
Posts: 5
Joined: Wed Sep 28, 2011 4:45 am

Re: problem with streamed rigid body data from Area

Post by rranon »

I'm doing some more precise test tomorrow and also explore the options suggested by Don (thanks!).

Mike, if you read this meanwhile and have some time to answer, can you please give me some more details on the differences between tracking tools and arena as far as tracking and streaming rigid body data are concerned?

Thanks a lot,

Roberto
rranon
Posts: 5
Joined: Wed Sep 28, 2011 4:45 am

Re: problem with streamed rigid body data from Area

Post by rranon »

Ok, I've made some more tests.

As reported by Arena, position data oscillate by about 1 millimeter in every axis, but it depends on the position in the tracking volume. In some positions I don't see any oscillation.

Orientation data oscillate by about .2 degrees, this seems not to vary by choosing different positions in the tracked volume.

However, I don't think this is the real issue. My problem is that, with an otherwise identical configuration, our application works with Tracking tools but not with Arena, where I see that:
(i) axes are flipped (e.g. rotating around Y produces the opposite rotation)
(ii) the geometry to which the rigid body data is applied oscillates a lot (much more than 1 millimiter).

With tracking tools, everything works fine.

So, there must be some difference between how Arena and Tracking Tools track or stream the same rigid body data, I guess?
rranon
Posts: 5
Joined: Wed Sep 28, 2011 4:45 am

Re: problem with streamed rigid body data from Area

Post by rranon »

I have found the source of our problems.

It looks like Arena streams data in millimiters, while we were streaming meters with Tracking tools (and assumed that in our application). So, a small oscillation in the range of half a millimiter was considered in the range of half a meter.

Moreover, given some position and orientation of the rigid body in the tracked volume, Arena and Tracking tools, apart from units difference, do not stream the same position/orientation data, but:

Pos_X_Arena = -Pos_X_tracking_tools
Pos_Z_Arena = -Pos_Z_tracking_tools

Quat_Y_Arena = - Quat_Y_tracking_tools

For the first problem: is there a way to configure the units of measurement used by Arena?

For the second, the behavior of Arena seems counter-intuitive, because moving in the tracked space along +Z (as defined, when calibrating, with the L ground object) decreases the reported Z position. Again, could this be due to some parameter setting, or error in the calibration procedure? The rigid body was defined with its local + Z axis in the same direction of the global +Z axis (or more precisely, the local coordinate space was aligned with the global one).

Last, I could not find any written information about these things. Maybe I did not search very carefully, but I think the absence of a manual for Arena is really disappointing, considering also that I did not receive any support in this forum besides the message from Mike 5 days ago - which moreover did not suggest any remedy or potential problems to investigate.
NaturalPoint - Mike
Posts: 1896
Joined: Tue Feb 01, 2011 8:41 am
Location: Corvallis, OR

Re: problem with streamed rigid body data from Area

Post by NaturalPoint - Mike »

As Arena and Tracking Tools evolved separately for different purposes, there are minor differences such as streaming mm vs m that need to be accounted for. Admittedly, at current there is a rather disappointing lack of documentation which is something that is being worked on. Our next version of Tracking Tools should be the first to release with a comprehensive user's manual, and Arena will be following suit shortly.

Arena and Tracking Tools both default to streaming right handed coordinate planes, so I find the unintentional translation a bit odd. What are your streaming options set to in Arena? Are you using a right or left handed coordinate plane in Tracking Tools (edit menu, tracking tools options, options tab)?

The arena issue you're experiencing could be due to a ground plane error. When looking at the ground plane, the longer of the two arms is positive z, the shorter is negative x, and the side of the ground plane with the markers (up) implies positive y. When the ground plane is set, before moving the ground plane, the marker that sets the right angle should be at the origin, the marker on the longer side should be at (0, 0, .4), and the marker on the shorter side should be at (.2, 0, 0). You can check this by clicking on individual markers on the ground plane screen and looking at the coordinates in the top left corner of the 3d perspective view pane.
rranon
Posts: 5
Joined: Wed Sep 28, 2011 4:45 am

Re: problem with streamed rigid body data from Area

Post by rranon »

Thanks for the reply.

We are using right handed coordinates in both Arena and Tracking Tools. In Arena, the other streaming options are to stream rigid body data - all the others are unchecked.

Thanks for the info on the ground plane, I will check the reported position of the markers in the ground plane.

Generally speaking, is Tracking Tools more precise / reliable when tracking rigid bodies, or the two software, with identical cameras configuration, should perform in the same way?
NaturalPoint - Mike
Posts: 1896
Joined: Tue Feb 01, 2011 8:41 am
Location: Corvallis, OR

Re: problem with streamed rigid body data from Area

Post by NaturalPoint - Mike »

Precision and reliability are the same, however there are significant feature set differences between rigid body tracking and full body tracking that set the two product lines apart. Additionally, as data needs generally differ between the two products, export format is different (Tracking Tools is capable of dumping data to CSV, Arena supports BVH and FBX, both support C3D).
Post Reply