FYI

> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] Behalf Of Robert Osfield
> Sent: Thursday, April 01, 2004 4:50 PM
> To: [EMAIL PROTECTED]
> Subject: [osg-user]Handling whole earth databases
> 
> 
> Hi All,
> 
> As I write this, my other machine is churning away generating a PagedLOD 
> database with about a billion polygons.  Its only a chunk of the earth, but 
> its wide enough area to see the curvature of the earth clearly when you a 
> zoomed out.  With such wide area models a number of issues arise, but first 
> I'd like to open out two of these issue for discuccsion :
> 
>    1) management of near and far clipping planes as one moves for a distant
>        view of the the whole earth to a local view with objects visible within
>        a meter away from the viewer.
> 
>   2) how to adjust the local up vector used in drive manipulators and flight
>       manipulators & physics models.
> 
> I'd like feedback on the community on my own thoughts below, and any funky 
> suggestions that might save us all some time :-)
> 
> --
> 
> My first thoughts on 1) near and far clipping are:
> 
> We basically need to a class that takes as input of the current position in 
> world coordinates, and output a new near and far value, which all sounds 
> rather easy... except we need to have need formula for computing the near and 
> far from the current position, and a framework for attaching this class to 
> the scene graph, and for the cull traversal to be able to use this.
> 
> As for the formula I have been thinking about having an ellipsode model for 
> computing the local height of the control point above the ellipsode and from 
> this compute the near and far value to use in some "magical" way.  
> 
> Another addition to this would be the automatic control of glFog (via 
> osg::Fog) to make the far distance gradually disappear before the actual far 
> plane cuts in.  As you get lower and the far planes moves inward, so does the 
> fog so you never see a line of terrain disappearing into the clipping plane.  
> 
> The framework might be as simple as just a cull callback attached to a node 
> above the world with a reference to a Fog object which decorates the world.  
> Only once I get to implementing it will I know for sure.
> 
> Thoughts?  Have others done similar things before?  Any ideas of your own?
> 
> --
> 
> My first thoughts on 2) adjusting for a local up are:
> 
> Up in our world is generally defined by us as away from the earth, or to be 
> more exact orthogonal to surface of the earth (ignore local gradients), or 
> perhaps better as the axis that gravity acts.  Since objects most physical 
> models use gravity as part of the computation it seems to make sense to 
> create an local gravity vector.
> 
> We are now back to having a class which takes as input a control position and 
> outputs a vector, with possibly the length of the vector the strength of 
> gravity which could neatly allow us to have a model for other planets.  The 
> orientation of gravity vector could either just be modelled on sphere, or on 
> an ellipsode, or perhaps even allow for local variations in the gravitational 
> fields (this is the type of stuff that virtual functions were invented for:-)
> 
> Once we have this class we are back to the issue of a framework for physics 
> models/manipulators accessing the local orientation.  For osgGA and other 
> codes to be able to access this data we'll need a standard framework, most 
> probably in the core OSG itself otherwise the dependencies list will just 
> spiral away.  osg::GravitationField anyone :-)
> 
> --
> 
> So I'd love your feedback, is this stuff you've been thinking about before, 
> and have already tackled and solved?  Does the above look promising, or fall 
> into any traps that need to be avoided.
> 
> Cheers,
> Robert. 
>      
> 
> _______________________________________________
> osg-user mailing list
> [EMAIL PROTECTED]
> http://dburns.dhs.org/mailman/listinfo/osg-user
> 

_______________________________________________
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel

Reply via email to