Re:Show HN: 3D sensing SDK for iOS – Produces point cloud an...
by sytelus on Monday September 09, 2013 @10:44PM
(#6357819)
Accelerators are worthless to measure distances even if they were super precise. This is because you have to do double integration to get distance and the errors only accumulates. My guess is that they have used algorithms like SIFT to track points in space and estimate only small relative distances from sensors (gyro + accelerators). This is however is very cool. They should make this as an app that can construct 3D model (i.e. turning iPhone in to 3D scanner) and ability to send the model for printing.
Re:Show HN: 3D sensing SDK for iOS – Produces point cloud an...
by yetanotherphd on Monday September 09, 2013 @10:56PM
(#6357863)
>Accelerators are worthless to measure distances even if they were super precise. This is because you have to do double integration to get distance and the errors only accumulates.
I never said that a distance scale was obtained by applying double integration to the accelerometer output. I only said that in order to measure absolute distances, as opposed to relative distances, it is necessary to have an accelerometer, since no other data provides an absolute scale. See the other reply by one of the founders for the details.
Re:Show HN: 3D sensing SDK for iOS – Produces point cloud an...
by eaglej on Monday September 09, 2013 @07:57PM
(#6357129)
You've explained it well! And you're exactly right, thanks to the accelerometer we get distances in real-world units. The trick is very closely integrating computer vision and inertial sensing. Images provide an external reference that can clean up the noise from the accelerometer, and the accelerometer provides absolute scale which you can't get from images alone. (I'm one of the founders, BTW.)
Re:Show HN: 3D sensing SDK for iOS – Produces point cloud an...
by benhirashima on Monday September 09, 2013 @07:34PM
(#6357035)
Yep, that's pretty accurate. I'll let our PhD guy come explain a bit more himself. Sorry, we're a bit busy with our exhibit at Disrupt, bit we'll be able to answer more questions later.
Because the device has an accelerometer, it is even able to extract distances, not just relative distances. I'm actually surprised by this as I always assumed the accelerometer was too noisy to be of use for this.
I tried to do a similar thing myself, but the problem is technically very difficult. While a lot of research has been done on structure from motion, actually packaging it into a usable API is a big task