On 20/09/2008, at 8:08 AM, Mike Liebhold wrote:
> wondering what the minimum metadata is required to use a random
> photograph as a location sensing template: camera position in three
> dimensions, compass direction, field of view, corner locations, pitch

Just need an approximate location and the SIFT descriptor for the  
object viewed from that location. Then a photo that is taken from that  
approximate location may or may not have the same SIFT descriptor  
(depending on visibility), which is a desirable result anyway.

That does mean that the same object will have different SIFT  
descriptors as the location changes, so that's the unique part that a  
location DNS would need to handle.

--
   [EMAIL PROTECTED]


_______________________________________________
Geowanking mailing list
[email protected]
http://lists.burri.to/mailman/listinfo/geowanking

Reply via email to