We may need to use multiple sensors on one object/in one room. Before we get more sensors… does anyone have experience with using multiple Structures in close proximity? How’s the interference between them?
I’m going to go out on a limb without really knowing, and say that my assumption is that it wouldn’t work. Unless the sensors are somehow in sync, I don’t imagine there is any way they could distinguish their IR pattern from any other sensors. It would probably just throw the measurements out of whack. I could be wrong- I’ve seen multiple Kinect setups but I’m not sure how this problem was solved.
Personally after a weekend of scanning sculptures as an experiment in France, I have found it would be really handy to be able to scan not only with the ipad (and preferably an Iphone) but to be able to put a sensor on a “Selfie” stick to get scan data beyond ones arms length reach.
I went ahead and tested this for you. In short, you may notice an impact in individual depth frame quality / reliability (noise, depth holes, pixels with random depth, etc), which is going to depend a lot on your particular setup and target.
I tried to make this as pessimistic of a case as possible - these sensors are right next to each other and are projecting their patterns over essentially all of each others’ FOV. Your application might not be so hostile, but YMMV.
The noise/shadows/etc seen here are very transient - I’m still able to get good scans of objects with more than 1 sensor pointed at them.
Side note - if your application isn’t fine with possible reductions in quality, you might be able to get away with turning on/off streaming on your devices in sequence (which will shut down sensors’ projectors when not in use).
@dennison - loving the selfie stick idea. =]
Thanks for the photos. We did some tests with Structure and Kinect v1 here, and the interference looks similar to yours.
Structure with Kinect on
The tripod is pretty challenging I suppose, since it is metallic/reflective.
Also some Kinect shots…
Kinect with Structure on
BTW, here’s a good selfie stick implementation. It’d likely support light field capture, too.
That god Selphie stick is not only hilarious, it’s actually not a bad idea. You could indeed do light field capture like that.
I want to use 6 devices to see a character placed in a pre-scanned part of a room. Will this character be viewable by all 6, will character shudder, and/or will the character shut down and sensor’s SDK throw an error message?
Check out our latest Youtube video for a test running Occipital’s Bridge Engine on multiple devices in the same area.
Multiple structured light IR patterns and activity of people walking around in the environment did not prevent relocalization & tracking of an augmented reality scene.
Hope you enjoy!
NOTE: This is running Occipital’s Bridge Engine, built on top of Structure SDK. Bridge Engine is built for mixed/augmented reality experiences.
This is very impressive!