Researchers use Kinect to increase 3D scanning accuracy by 1000x

Chris Scott Barr - Dec 21, 2015
0
Researchers use Kinect to increase 3D scanning accuracy by 1000x

It’s hard to deny the fact that the Kinect has all but failed as a gaming device. Despite their initial decisions, Microsoft stopped including the device with all Xbox One bundles, which fractured the install base. And with that, fewer and fewer developers felt the need or desire to implement its features into their games. However, the device is still being used to make breakthroughs in other, non-gaming areas.

Earlier this month we told you how the Kinect is being used to take better X-rays. But now it would seem that a different group of researchers is using the device for 3D mapping, and it could be a real game-changer. How much of a game-changer? How does 1000x better mapping strike you?

The group of researchers at MIT have developed a new system which they call Polarized 3D. Essentially, they had been looking at a way to use polarized light to map a 3D surface. While it is possible to calculate such things, due to the amount of raw data that must be computed, it’s simply not a viable option. Enter the Kinect.

Something that the Kinect does very well is measure the depth of objects in front of it. And when you take the information gathered from measuring polarized light, and add in the information about an object’s depth from a Kinect, you apparently get a very detailed 3D image.

The researchers used a Kinect with an ordinary polarizing photographic lens in front of it. They then captured three images of an object. After running the captured images through their custom algorithm, they were able to produce stunningly accurate 3D renderings. The renderings showed that they can accurately resolve features in the range of micrometers.

The team put their system up against a more traditional laser scanner, and not even it could beat out their results. While this research sounds amazing, we’re probably some time away from it being put to any real commercial use. However, the team is planning on releasing their source code, so that anyone can attempt to put this information to use on their own.

Source: MIT


Must Read Bits & Bytes