forums.naturalpoint.com

I've recently purchased and setup the Optitrack system with 8 cameras (R2), and have repeated a few sessions of setting up and capturing.

Sometimes i'm getting really great results, but at the Arena stage I can see markers being lost pretty consistently (going grey), which leads to skeleton areas being lost from tracking - going red and being lost completely - which really ruins the usefulness of the data in many cases. Mostly it's the top torso segment or the upper arms.

I'm putting the lower cameras very close to ground level on their own mini-tripods and upper cameras are mounted to top the tops of the lightstands. This could be the source of my problem - though I tried raising the lower cams to 2 foot for one trial and got worse results than on ground. I suppose if this is definitely not recommended my next step should be to rig the lower cams as recommended - but I've read others doing the same as me here.

Currently I'm really only getting short periods of time when everything tracks solidly and accurately.

Staging area is around 13 x 16 feet. 8 foot hight. Gives a good volume to move around in.

I can pretty consistently get all 'excellent' results in calibration runs, maybe a few 'great' mixed in.

Two powered USB hubs are being used, and I connect one side of the room to one, the other side to the other. 4 each. The cameras read 100fps in Arena. I'm not sure though if each hub is running on it's own controller on the motherboard - although i have multiple USB controllers on the MB, every USB device appears on the same tree in device manager and arena. Not the same branch, but the same tree root. The good calibration results hint to me though that my USB throughput isn't the problem.

I run the analog cables to synch the cameras. I set the camera order up the same every time going off the serial numbers.

I aim cameras fairly much according to instructions, just putting the lower cams tilted up a bit to avoid shining into the ground.

I block out the opposite ir light rings in software, and a few minor reflections in our studio (in software). Blocking areas seem concentrated mid-frame due to - well, cameras being opposite. I've experimented with the various exposure / threshold / intensity parameters - markers however seem to be illuminated great, and i only need to block the ir rings and some of their reflections. I think I'm about to start experimenting with some different camera arrangements such that these blocking areas aren't all concentrated mid-frame.

I suit up myself or an actor, following things such as back marker level with shoulders in T-pose, shoulder markers lower down from the joints, hips level, etc.

As said - calibration results are excellent. Though when it's time to match the skeleton with a T-pose - I can see immediately the system is consistently losing track of markers - because all 34 of them have to be there to lock in the skeleton to the T-pose in arena, and this in itself usually requires a few tries.

Capturing after that is a bit hit-and miss. The system has a lot of trouble 'locking in' to a solid skeleton and we really have to ease motion and our timing of things to get any good capture.

I'm running the latest arena software 1.5 beta 2.

I'm sure i'll get there - I'm getting periodic excellent capture. Lower body is excellent. No matter what tweaks I make to maker placement I can't seem to stop the top-torso and upper arms jittering and losing tracking completely.

Thanks for any replies.
-Kristian

OK, a few things to look at -

1 - The upper cameras at 8' are ok, but the lower cameras should be approx. 2-2.5' below the upper cameras, not on the floor.

2 - Try rotating the lower cameras 90 degrees, for your smaller campture area.

3 - When aiming all of the cameras, attach the wand to the ground plane, and set the wand single marker to waist high. and aim all of the cameras at that marker.

Also if you can send some sample calibration, wanding, t-pose and capture data to support, we can look at a few other thinga as well.

Thanks for the quick reply Vincent - I set things up again today with upper cams 2.3' below the lower ones, and aimed for centre at waist height, also rotated lower cams 90 degrees. So took on all 3 of your suggestions.

Solved the problem.

Solid skeleton now - markers stay tracked full body, pretty much continuously while in the capture volume in reasonable posture. Blocking areas are now top-frame rather than concentrated in the center of frame.

Glad to hear it. Let me know if you have anymore issues.

I've come across a problem that is really my fault but I'd like to send you some data files if you wish to examine this.

As I mentioned I'm getting solid capture with your recommended camera arrangements, and captured a scene full of data last night which went great.

Problem is - I seemed to have saved over the Arena project file as an empty file - so lost the .mcp project file.

Question is - can I feasibly build the project back from the saved data components? The wanding, calibration result, skeleton file, and Tpose data I have saved. Also of course I have all the captured takes saved.

If yes - could it be a problem that Arena is crashing when attempting to load in the calibration result, and giving an error 'An invalid argument was encountered' when trying to load in the skeleton file ?

emailing to support the 4 setup files, one capture sample.

Thanks for any help with the project re-build issue, and also pointers on how I could improve data quality based on those samples - if you can use them without the project file. Also I'm keen on looking into some features of Arena such as Camera Processing mode, Tracking Parameters, and Point Cloud settings to imrove data quality and tracking further. Is there any documentation on these things as it relates to mocap operation ?

Answered via email -

For the calibration file, I just reloaded the wanding data, and reran the calibration itself, which created a new calibration, which doesn't crash.

Most of our documentation is contained in our tutorial videos.

http://www.naturalpoint.com/optitrack/p ... ideos.html

Thanks, yes I can re-build the project again here.
I loaded the calibration points and ran calculation as you suggest. I had to remember to use 120 samples rather than 100 here because that meant the difference between mostly excellent results vs mostly Good / Fair for this data.

To rebuild the skeleton I opened the wizard, loaded in the T-pose data (bypassing the edit existing template screen), then fit T-pose to this data which enabled me so save a new .skl file.

Somehow the software worked out how to orient the groundplane again, correctly.

Very good to know how to rebuild the project.

As long as youy have fully covered the capture volume, with the wanding, I wouldn't put too much stock getting either excellents or goods, as long as the final data gives good results.