Kinect shortcut

Have you try using the tryplex qc kinect macros?

No. I was guessing I would not pass the cactus with the kinect hanging far overhead? I´ll have a look, thanks,-

You don't have to do a pose with newer versions of OpenNI, but it can't track a skeleton from above either.

I don't think that any other blob tracking methods that are particularly more robust than the tuioKinect app's is. The near/far threshold are pretty standard. Your best bet is to try to bring the cameras lower, and/or minimize any light reflections from the floor. You might try adding some other filtering on the ofx side (via changing the src), like maybe some dilation to help get rid of smaller "noise" pixels.

Thanks, I saw that about OpenNI and no pose coming, but seems not to be implemented in something easily accessible for me yet. Not trying to track a skeleton from above (I have an MRI scanner for that..), just the general whereabouts of the actors,

I´ll have a look at the Tuio, but the cameras can not go lower.

ofx= Open financial parcer src=Proto-oncogene tyrosine-protein kinase? Nevermind, I´ll get back after looking...

I don't think I ever sorted the blob centroid thing properly as I was experimenting before open ni was available to the public. if I remember correctly I did make a plugin that calculated the centroid much faster than I was able to do witht the JavaScript patch. if you make the depth image very small things work better as there not as many pixels to track.

those comps should work fine for one actor. the comps are all qc based and do not rely on the robust open cv blob tracking that caresuelo and tuio use. I haven't put much thought into the qc based blob tracking approach in some time. at one point I had the caresuelo blob tracker working with the Kinect, I will see if I can dig up that comp. you might want to mess with the kineme open cv plugins and the depth image ?

at first thought I was thinking you should be using the Kinect front facing and tracking skeletons but I think 6 people maybe to much. I have got three skeletons going before. either way at five meters out your beyond skeleton tracking. so your on the right path using the depth image blob tracking approach. you may be able to syphon out the depth image and feed it to a multi touch blob tracker then receive the touch data via tuio in qc. also I think vades Kinect plugin provides more pixel data than the kineme plugin so that might be a better starting point ? either way with my centroid tracker I was using a very small image, I think smaller than 100 x 100.

Have a look at kinectA for mac & win – maybe it suites well into your project. I think it is worth to look at, because it combines three elemental parts of tracking: http://kinecta.mihoo.de/what-is-kinecta

KinectA is a relative new motion tracking application (beta release at 07-2012), which combines hand, skeleton and object tracking. It make use of ofxOpenCV and ofxOSC. The distance to the tracking object is controllable over the GUI.

Thanks for your propositions dust, I´m going for the top camera version, stage depth is 9 meters...

The KinektA works nicely, and is easily finetuned. Downside is that it eats 4 times the CPU of the Tuio and unfortunately it seem unstable on snow. I´ve had three crashes in a half hour. so -back to the Tuio for now.

Thanks for telling anyway 3Lab vj

Preview	Attachment	Size
	Skjermbilde 2012-10-29 kl. 22.17.08.png	30.97 KB

Actually I´m lost in how to get more than xy`s out of the TUIOKinect.

The TuioClient plugin outputs TuioCursors where i get ID, x,y, xspeed, yspeed, maccel?? , but no size, height etc, and the TuioObject output seems dead.

I´ve got qcOSC recognizing more inputs, but no data flowing... I was hoping for z and size.