Spine still bends oddly

Greenlaw · **Posted:** Tue Feb 26, 2013 9:45 am

This might have more to do with a problem with my footage but since I'm not sure yet, here's what I'm seeing at the moment after Calibration using build 147.

Attachment:

2013-02-26_083510.jpg [ 25.1 KiB | Viewed 18032 times ]

I saw something similar to this yesterday the first time I ran Calibration using build 146 but the second time I ran Calibration, the results looked more correct--though still slightly off-level obviously.

I just tried calibrating again in 147 and this time it looks better, though still off-level by a couple of degrees--just like yesterday's results.

I'm almost convinced now that this is just bad footage. I'm going to try recording new video with different camera placement now.

G.

Greenlaw · **Posted:** Tue Feb 26, 2013 9:52 am

Just wondering, could the brightness of the white board affect calibration? It's not a shiny surface but maybe the pure whiteness makes the IR difficult to detect?

I'll see if I can find a board with a color or darker surface.

G.

Greenlaw · **Posted:** Tue Feb 26, 2013 10:40 am

Yes, I think I had bad Kinect sensor positioning last time--they must have been set too wide for the 90 degree configuration.

It's still also possible that the brightness of the board does affect IR detection. I tried using the same white foam core and a flattened brown cardboard box for calibration, and the brown box test has a lot more 'green' trackers.

I still need to bring in an action video and see how things are oriented with the adjusted Kinect sensor positions. More to come.

Greenlaw · **Posted:** Tue Feb 26, 2013 10:54 am

Confirmed--it must have been bad footage. The point cloud with the new footage is standing perfectly straight up, no tilt. Yay!

Attachment:

2013-02-26_094822.jpg [ 30.95 KiB | Viewed 18031 times ]

However, why does the view from camera 2 look like this?:

Attachment:

2013-02-26_094844.jpg [ 31.81 KiB | Viewed 18031 times ]

Here's what the view from camera 2 looked like at calibration:

Attachment:

2013-02-26_093254.jpg [ 58.27 KiB | Viewed 18031 times ]

I don't know yet if this is anything to be concerned about. Is this view correct and the camera icon is just offset? Or is the calibrated position actually wrong? I'm not sure--please advise.

Thanks for putting up with my nonsense. :)

G.

Edit: I previously wrote 'Recorder' where I obviously meant 'calibration'. Sorry for the confusion.

Greenlaw · **Posted:** Tue Feb 26, 2013 11:37 am

Re-recorded calibration and footage. Holding the flattened brown box does worked even better when I held it about a foot lower than usual--nearly all frames have green trackers now.

However, view from Camera 2 is offset again, although the offset is different this time. I'm going assume this is a bug in build 147 since the offset issue keeps occurring:

Attachment:

2013-02-26_103354.jpg [ 68.86 KiB | Viewed 18030 times ]

BTW, this time, I was using the 180 degree configuration.

G.

Greenlaw · **Posted:** Tue Feb 26, 2013 11:44 am

The white board actually tracked much better this time--like the brown board, I held it about a foot lower than my usual position. My guess is that this has to do the angle in which the IR is detected.

Anyway, I'm getting ALL green markers with the white board now. (371 out of 373 frames actually, but close enough to perfection.) :)

Curiously, the view from camera 2 looks good now too. Puzzling.

Attachment:

2013-02-26_104335.jpg [ 74.9 KiB | Viewed 18029 times ]

Oh, well--I'm gonna go with this then.

G.

Greenlaw · **Posted:** Tue Feb 26, 2013 12:20 pm

Ugh! What appeared to be a perfect calibration actually looks like this:

Attachment:

2013-02-26_111811.jpg [ 60.48 KiB | Viewed 18029 times ]

Another user reported an issue like like this yesterday (look for thread regarding 'Stacked Cameras'.) Could this be a new bug when using the 180 degree configuration?

G.

Greenlaw · **Posted:** Tue Feb 26, 2013 12:26 pm

Looked at the calibration with the brown board and I see the same 'stacked cameras' issue.

Could the problem be that the Kinect sensors are positioned too close to 180 degrees? They're actually positioned less the 180 apart but it's still pretty close.

Will try moving them to even farther from 180 degrees. If that doesn't work, I'll go back to using the 90 degree configuration.

G.

Greenlaw · **Posted:** Tue Feb 26, 2013 12:39 pm

180 degree calibration is definitely bugged. Here's the result after positioning the sensors to 160 degrees apart:

Attachment:

2013-02-26_114801.jpg [ 50.37 KiB | Viewed 18027 times ]

I'm guessing coordinates are incorrectly flipped somewhere.

Going back to 90 degrees now.

G.

Greenlaw · **Posted:** Tue Feb 26, 2013 12:57 pm

Before moving on to 90 degrees, I decided to try one more thing.

I notice that when the calibration scene first opens, the cameras are stacked, so I swung the camera to the other side using Orbit mode, and then initiated calibration. This time it worked.

Attachment:

2013-02-26_115706.jpg [ 43.82 KiB | Viewed 18027 times ]

Apparently the new automatic 180 degree configuration detection feature is not working in 147 and Mocap Studio is assuming that all calibration is 90 degrees.

Anway, with this workaround, it looks like we're back in business. :)

G.

Spine still bends oddly

Who is online