PDA

View Full Version : 3D Camera



waldronate
08-19-2009, 03:22 AM
Today I saw an article where Fujifilm is coming out with a 3D camera. http://www.fujifilm.com/products/3d/camera/finepix_real3dw1/ has the specs except for price. Some folks are thinking around $600. The fun feature is the 3D viewing screen in addition to the 3D image taking.

I have tried and tried to think of a good way to use this for mapping purposes but I just quite can't come up with one. Anyone else?

Steel General
08-19-2009, 08:10 AM
I'm sure there's a practical mapping use for this, but I have no clue what it might be.

I can barely use the camera on my cell phone. :D

Alfar
08-19-2009, 08:59 AM
I suppose with information about the state of the camera (angle of lens, zoom factor etc), having two pictures of the same area at the exact same time and with a known distance between them, it'd be possible to calculate distance and direction to any given object in the image. Map that info onto a top down view of the area, and you have a map of sorts, I suppose? ;)

Korash
08-19-2009, 01:31 PM
I wonder if the 3D image could easily be turned into a hight field? Like a one picture Steriotopic sort of thing.

ravells
08-19-2009, 04:10 PM
I'm sure there's a practical mapping use for this, but I have no clue what it might be.

I can barely use the camera on my cell phone. :D

I'm glad I'm not the only one. I find my SLR easier than the cell phone camera!!

Alfar
08-19-2009, 04:22 PM
I wonder if the 3D image could easily be turned into a hight field? Like a one picture Steriotopic sort of thing.

Well, not easily, since you don't really gain any information on a per-pixel level.

The only information you can get from these 3d images that you can't from a regular 2d image is that when you figure out which two pixels depict the same spot, you can calculate the distance between that spot and the camera. Unfortunately, that doesn't give you any height information, though I suppose if you turned the camera on it's side, you could get some info out of it.

Also note that if you take a picture of a hill, the picture can't magically tell you anything about what's behind that hill, so even if you could make a height map with it, you'd get nothing for anything that's hidden behind other stuff.

Of course, if you were to take a picture vertically down from somewhere high, you could get height-map info. Like, say, a satelite... hmmm. ;)

Redrobes
08-19-2009, 05:14 PM
Theres some great movies on Devebec's site where he took stereo images of a forest path and gets 3D world data out of them. He uses it to interpolate the tweens when moving forward so that as he walks on a path taking a photo every 10 or 20ft he can then make a full frame rate movie of moving down the path.

http://www.debevec.org/Immersion/
http://www.naimark.net/projects/dim.html

Alfar, your right in that you have to compare the two images to get the Z info along the camera lens axis. If you took the photo high up then the two images would essentially be the same so you would not get any appreciable difference enough to get any accurate Z at all.

But suppose you already had a height map which was super low res like the ones you can get free. Say a 100m point sampled one. On the ground that translates into some pretty triangular blocky landscape. From knowing this as a starting point you can then take photos of an area and add extra samples in. Sure you need to take many and you need to known the location from where you took it but from having the rough height map you can use that to correlate the Z info to determine the direction of the camera and use GPS to get the location which some cameras have built in now. So with many, you could get a high res height map of terrain with all the color photos needed to texture it a) from above and b) from the sides too in the cases of trees etc.

Ambitious I would agree.

Where I think I might be most interested in this is for scanning 3D objects using structured light patterns projected onto it.