Follow-up on Multiple Kinects
January 20th, 2011 | 
Author:
Video summarizing my current opinions on how one could have multiple Kinect devices capturing in the same space at the same time, summarizing the ongoing discussion in the comment threads.
Also a shameless plug for my “official” YouTube channel, spelunkerucd. Check it out if you want to see the kind of stuff that might come to your living rooms in a few years or so.
http://www.youtube.com/user/spelunkerucd
Duration : 0:5:47
[youtube YB90t5Bssf8]
Posted in Sales Follow Up | 
Tags: 3D, capture, interference, Kinect, merging, multiple, multiplexing, video, Xbox, yt3d:enable=true, yt3d:swap=true
looks like the …
looks like the money for nothing video
is this guy dutch?
is this guy dutch?
If he could find a …
If he could find a way to use several kinects so the pc could reconise a full 3d feed of a house or somthing it would be the best security and he would be like a billionareeeeeeeeee!!
hahaahh you’re at …
hahaahh you’re at work
I think the …
I think the polarzation filter idea should work and would be interesting to test out.. Ideally to create a 3d object with 2d planes you would need 3 views, like front, back and side. With the kinect giving us depth however, we would only need the front and back. For any sides we can’t see use the depth between the two points to interpolate the colours to fill in the void.
If the connect uses …
If the connect uses lasers to project the infra-red, then the time multiplex is very possible. Look into phased array RADAR technology the U.S. Navy has developed for the Aegis platform.
@jonkam10 YouTube …
@jonkam10 YouTube messed up the stereo on this one. It looked perfectly fine when I created it.
Regarding the game character: since the Kinect generates a real-time 3D model of an environment, it’s straightforward to just drop additional CGI models into that. You just need to measure the 3D point where you want the model to appear.
why is this dude so …
why is this dude so nerdy.. jk this is dope.
sorry, but red/cyan …
sorry, but red/cyan should be way more close together! I like your doom character 3d video, I would really like to understand how that works! I used motion tracking and stuff but didnt digg the results so far. Vielleicht sprichst du ja auch deutsch? peace
REAL Free energy …
REAL Free energy technology exists!But the Oil coporations life depends on covering this up,Find the real deal, a free energy device at LT-MAGNET-MOTORdotCOM ,Let the revolution begin!
i like germans
i like germans
where can i get …
where can i get software to run a kinect on my pc
wiah im having a …
wiah im having a hippy momment
Could you not have …
Could you not have 3 Kinects w/o multiplexing if you put the third one above the subject looking downward with a polarizing lens?
@alexutubeutube Yes …
@alexutubeutube Yes, we’ve known that for a few days now. Check “2 Kinects 1 Box”, /watch?v=5-w7UXCAUJE
Using multiple …
Using multiple kinects adds noise to the depth measurement but can still be used for advanced processing.
Several kinects are used to detect and track people:
see video on youtube:
“Real-time People detection and tracking with multiple Kinect cameras”
smart people are …
smart people are awesome
this is awesome im …
this is awesome im buy kinect in december <333
realy kinect sucks …
realy kinect sucks if you use it like its meant to be used but if you do stuff like this with it its just really awesome!
These videos have …
These videos have been some of the most exciting developments I’ve ever witnessed in my life as it was still blooming. Thank you so much for your awesome work. Combined with some of the similar types of software recently and currently being developed, especially in robotics, this blows me away with what may be possible within just a year.
some one send him a …
some one send him a mirror.
@robvh2 Well the …
@robvh2 Well the point is that if you have multiple Kinect, they may actually do the combining already. I’m not sure how yields are for fabbing an IR lattice that’s the same _every single time_ so it would make sense that there’s some error tolerance involved here at least. As for resolution, I don’t know that the CCD in question has been resolution tested (yet), and it may change in any event. The CCD and lens are going to be really important for overall sensitivity, no?
@WyattEpp Other …
@WyattEpp Other thoughts realted to density. You only get a decent 3d model if you have a dense enough point cloud. So that determines how far away the Kinect can “feel”. I think the near limit is determined by the resolution of the camera. As the subject gets closer to the Kinect, the IR dots get closer together. Once the dot density exceeds the camera’s resolution, then it can no longer differentiatie the dots and the Kinect would see only a blob of light. Make sense?
@WyattEpp If you …
@WyattEpp If you can generate more than one cloud, I see no reason why you couldn’t combine them into a single cloud of greater overall precision.
I suspect that the Kinect projects a pre-determined IR pattern. However, that might not be the case. For all we know, the wavelengths may be slightly different. That gives me an idea! Rather than using polarizing filters (or using the “poor man’s option” of tilting the 2nd Kinect) could you use some sort of IR filter to better differentiate?
@robvh2 @okreylos …
@robvh2 @okreylos What I really want to know right now, then, is if this works with more than one Kinect, can you basically get cheap linear increases in precision? (It’s ray-casting in the real world). This goes back to what I was asking earlier, about whether the hardware silently supports arbitrary patterns: doing so lets MS test/release revisions (patterns to improve accuracy, FOV tweaks, higher resolution through greater density, etc.) without having to redo the firmware. Thoughts?