Getting Started With Two Kinect Sensors


#1

Hi! Is there a getting started documentation regarding proper setup for how the sensors should be pointing at the figure doing the skeleton tracking?

I’m probably not going to get into mocap unless I can find out how to exactly take advantage of two sensors. I’ve seen the limitations of what a single kinect2 for windows can do on my computer. So basically my question is, if the sensors cannot tell if the tracked figure is facing front or back, then at which angles do the sensors need to be pointing at the figure so that I can get a more accurate read for where the limbs are?

I also read that the data for each sensor needs to be recorded separately and then combined later in software? Which software merges that kind of data together accurately? Then what is the point of buying two kinect2 sensors then, Just so you don’t have to move the single sensor around?

Thanks for any hints on where I can get these answers for using two kinects together properly.


#2

Two sensors will conflict with each other. The Kinect projects an IR dot pattern in front of them and observes the distortions in these patterns to calculate depth. If two Kinects are pointed in the same direction, their patterns will overlap and conflict. There exists an academic paper about attaching a servo on the Kinect’s body to vibrate it at some frequency, and which point the IR dot pattern conflict can be eased a bit.

Two Kinects can be used to get a larger area in view by point them in different directions, but they won’t work tracking one person at the same time.

Other than that, to get ideal conditions:

  • Avoid sunlight or other IR sources (close the curtains)
  • Make sure the full body is seen (including legs)
  • Position the Kinect at a height of about 1-2 meters
  • Make sure the floor material does not reflect too much IR light