Hi all,
kinect is wonderful and cheap , but it has its limitations.
what other devices and / or technologies are there to enable
-more accurate tracking?
-deeper depth tracking (>4m) and >7m ?
-perhaps quicker response from kinect (which is approx 60msec) ?
thank you!
I was playing with Kinect some years ago (with Flash / OpenNI).
First, i think the new version (Xbox One) of the device is more efficient, isn’t ? Have you try the Kinect 2.0 ?
Other way, you can try Xtion Pro http://www.asus.com/Multimedia/Xtion_PRO_LIVE/
or maybe Leap Motion if you wan’t to track hands.
Could you define your terms a bit? And could be a bit more specific about what you are trying to track (dancers? mosquitos? hands?). Unconstrained tracking (i.e. no markers, open space) is always going to be a bit more challenging …
These look interesting and apparently work with OpenNI - I bet the price reflects it as well 
I recently had good luck with D-imager cameras, which are time of flight cameras from panaosnic:
http://www2.panasonic.biz/es/densetsu/device/3DImageSensor/en/
they are not cheap, but I got one that works well in daylight, which is close to impossible with the kinect.
If you are on Windows, the time of flight camera in Intel’s Creative Interactive Gesture Camera Dev Kit (available commercially as the Creative Senze3D) might do the trick too.
hi all! thank you for your replies!
well the aim is to track dancers and objects on stage.
I have used kinect and can identify some weaknesses. For example , It is impossible to extract a noise free blob of a person leaning or standing in front of a wall (or laying on the ground).
So i’m looking at other cameras as well. Fotonic seems promising. I’ll check out D-imager too.
The asus xtion , primesense carmine seem of similar abilities to kinect.
bakercp i am not sure what you refer to with “unconstrained tracking”.
I should also note that I prefer to use linux … openNI OF wrapper doesn’t seem to work with it… so i’ll have to see what else is there.
By unconstrained, I mean markerless and without physical or gestural constraints in a given space. For example, if you know that the only thing you can possibly see in your see is a hand – or you only need to trigger an event when a dancer assumes a certain predefined position, then “tracking” with that constraint becomes a lot easier because you can make a lot of assumptions. That said, it sounds like the rest of the people on the thread understood your question better though and came through with more helpful replies
truth is, I would very much like to be able to track edges of the body (just hands, or just head, or just feet) when viewing the whole actor, but i guess this must be done on software level. and skeleton tracking in not an option for OF + linux , if I am not mistaken.
openNI works on linux and does skeleton traking
openNI wrapper with OF does not though, or have things changed. The documentation on ofxOpenNI does not state that…
@zach We are considering getting the D-Imager EKL3106 with the high lux rating to do blob tracking of people on an LED floor. Is that the one you have been using? In the spec they say that it works at 5 meters distance, but I am a bit worried about noise at that distance - what was your experience? /lars
Hey, have you tried to do some CV stuff with the rgb image? it might help a lot in the case that you describe, where performers are against a wall. a running background subtraction with the help of a dilated depth mask might work well.
check theo’s super nice kinect arm tracker https://github.com/ofTheo/kinectArmTracker
you can get a quite good idea about how to do it by analyzing it.
best!
Cool, I hadn’t thought about that. I will check the arm tracker out.