spatial mapping cameras

james_ashley · March 2016

The documentation says: "The camera used for scanning provides data within a 75-degree cone, from a minimum of 0.8 meters to a maximum of 3.1 meters distance from the camera. Real-world surfaces will only be scanned within this field of view. Note that these values are subject to change in future versions."

Does this mean only the 1 depth camera is used for spatial mapping?

Or are the 4 IR sensors used for this, too? Or are the IR sensors just for gesture tracking?

ChimeraScorn · March 2016

Spatial mapping can be thought of as a realtime analog to Kinect Fusion. Fusion involves two components, localization, and integration. Localization is basically figuring out where in the world you are. The various cameras in the device are used to absolutely nail localization, which is also needed for rendering holograms in the right place. The depth camera provides the depth data for the integration phase. There are a bunch of papers on Kinect Fusion if you're interested more in the spatial mapping capabilities.

ChimeraScorn · March 2016

Spatial mapping can be thought of as a realtime analog to Kinect Fusion. Fusion involves two components, localization, and integration. Localization is basically figuring out where in the world you are. The various cameras in the device are used to absolutely nail localization, which is also needed for rendering holograms in the right place. The depth camera provides the depth data for the integration phase. There are a bunch of papers on Kinect Fusion if you're interested more in the spatial mapping capabilities.

james_ashley · March 2016

Thanks @ChimeraScorn ,

I actually know Kinect Fusion really well. My confusion is really about the difference between the roles of the EW cameras and the depth camera. It's much clearer now. EW is for positional awareness. Depth for spatial mapping.

DaTruAndi · March 2016

I somehow hope that the additional cameras at one point will also contribute to the integration phase ... e.g.: http://research.microsoft.com/en-us/projects/monofusion/

gorgepir · May 2016

Just to see if I understood, does this mean that if I scan something from say 1 meter away, and then go closer to the object spatial mapping will no longer be able to recognize where I am?
Is there a way to play around with the 0.8 meter value?

Jimbohalo10 · May 2016

@DaTruAndi said:
I somehow hope that the additional cameras at one point will also contribute to the integration phase ... e.g.: http://research.microsoft.com/en-us/projects/monofusion/

There is another paper on how to create MonoFusion using a Microsoft LifeCam webcam.

Lots of complex math, but PDF makes interesting reading. There is no code, just describing potential functionality http://research.microsoft.com/apps/pubs/default.aspx?id=199618

Chop · October 2016

Are we able to do spartial mapping with kinect? i would like to accomplish the same as Hollolens with kinect. is it possible do you ahve some docs? what about hollolens SDK will it work? by just passing the kinect data as its pretty much the same?

james_ashley · November 2016

@Chop With the Kinect it was called Surface Reconstruction. Just a name change but same underlying idea. You generate meshes in both cases but the level of detail is probably going to be different. What are you trying to do?

spatial mapping cameras

Best Answer

Answers