Chris Goad schrieb:
Also, as Anselm mentioned, another hugely important case for AR is
multiple instancing. The notion of "global coordinates" of the
features of a manufactured item that has many instances makes no sense.
And an observation: For moving objects, the underlying reason that
local representation is sensible is that the local coordinates of
features are the invariant facts - facts that will remain true over
indeterminate motions in the future, unlike global coordinates however
well known at any particular time.
I agree that every object itself should have a representation in its own
origin, at least for editing them. You can get this by substracting just
the offsets from the absolute crs. That you are modeling this
environment object by object in its own space is clear.
But to render them on your device, you just have to know their absolute
coordinates.
A feature like a label for augmenting an object should be stored with a
semantic connection to the object (in real world) it is augmenting, and
not with a geometric coordiante in absolute space. This feature could
(or indeed should) have a geometry, and carry some information on how it
is positioned relative to the object, to be rendered. This geometry of
the feature should of course be stored in a "local" coordiante system
with its own origin, so that you can render it dependent of the view and
maybe to minpulate it as a user.
Look at the representation of a citymodels in CityGML. All geometries
belonging to a "parent" crs.
If precision, convenience are of no interest than yes a (x,y,z)
Geocentric CRS will work - but differences in typical large scale
applications will be very small.
@Ron: I don't get this... Why can't I represent an object in
submillimeter precision in that reference frame?
If your application recognizes (e.g.
image processing) some object in the environment - it would be normal to
specify things relative to the object (e.g. labels, measures on the
object etc) and these are more sensibly expressed in a local coordinate
system (typically rectilinear) - you may not know the accurate location
of the object itself (nor its orientation) - but you only care about its
approximate location (e.g. via GPS) and then the more precise location
of things relative to the object. It is not a question of just
differencing - since we will in fact know the local coordinates much
more accurately than the coordinates of the position of the object
relative to the earth.
sorry , I can't make sense of this in the moment...
regards,
Christian
_______________________________________________
Geowanking mailing list
[email protected]
http://geowanking.org/mailman/listinfo/geowanking_geowanking.org
