1

Problem

I have read that the Leap motion sensor is supposed to be capable of 0.01 mm measurements of hand gestures. I don't see how that is possible if the depth images are only 8-bit images (256 values at 0.01 mm per value would only yield like an inch of measurement range, which is clearly not the case). I've read through the API for the Leap Motion sensor, I don't see any configurable settings that would yield, say, a 16-bit depth image, which I would expect from a sensor claiming 0.01 mm measurements (see this site for that claim). Does anybody know a way to increase the bit-depth of the depth images from the Leap Motion sensor?

Context of problem

I am trying to use a Leap Motion sensor as a cheap 3D scanner. I want to measure the position and tilt of a 3D printer head with it. I think I can fake better resolution by analyzing many points across the surface, but if it is possible to get 16 bit resolution of the depth image, I would much prefer that.

I am not against using a different sensor if that is what's called for, but I am aware of other SO questions in that vein.

rexroni
  • 443
  • 4
  • 15

1 Answers1

3

The Leap Motion device doesn't use a depth image, so the color resolution of the images is irrelevant. (It uses stereo vision.) The API will give you a set of stereo images and a calibration map to remove lens distortion, but nothing else that would help you make a 3D scanner. To get the depth of a point in the scene, you would have to identify the same point in both images and triangulate. How accurately you can identify the corresponding image points would probably have the largest impact on the accuracy of your 3D reconstruction.

Charles Ward
  • 1,388
  • 1
  • 8
  • 13
  • So I had been told it was a structured light sensor, like a kinect. But it makes sense that the image I was pulling wasn't a depth image, just a regular IR image. It also makes sense now why there's two of them. But I'm not sure why my computer processor isn't going crazy if it's a stereo camera. Stereo vision takes a lot of processing, in my experience. This guy (http://blog.boulabiar.net/2012/07/unveiling-technology-behind-leapmotion.html) says that they're using some other technique based on different focal lengths, but he seems to be just guessing. Do you have any insight into that? – rexroni Oct 02 '15 at 18:40
  • 2
    That article is pure speculation. I only skimmed it, but the guesses didn't even seem close. You can look at our patent applications for how the device works (after all, sharing knowledge is one reason for patents to exist): https://community.leapmotion.com/t/leap-patent-application-and-other-references/717 – Charles Ward Oct 05 '15 at 17:23
  • 2
    This article offers a fairly brief overview of how the technology works: https://medium.com/@LeapMotion/how-does-the-leap-motion-controller-work-9503124bfa04. – leapmotion_alex Oct 06 '15 at 13:47