I am about to make an installation using 3 kinects to cover a larger area. Is the any experience in merging the depth maps from the kinects and ideally using all 3 kinects as one, both in regards to the depth map but being able to track skeletons across all 3 kinects so one person can move seamless from one kinect to the next.
Currently I am thinking that it must somehow be possible to do some sort of conversion from distances from the sensor to distances from plane of kinects.
With overlapping Kinects you have to deal with interference. Here are two interesting articles on how to overcome these problems by using a small vibrating motor:
Being able to track across the whole range would mean to calibrate the cameras accordingly. You could have a look at some of the projector calibration scenarios - e.g. using a checker board etc.
I think it should be possible to achieve this also without doing your own tracking, by comparing joint positions and knowing the cameras position / angle.
I think this is interesting too, if you wanna make a joint effort at a multi kinect module i would gladly help.
