google earth/virtual earth like thing
Author
I wonder how difficult it would be to create something like that. I mean, that basics seem to me that it wouldn't be hard at all.
1. You create a sphere, add some kb and mouse control
then comes the important question. Do you texture map one giant terabyte texture to the sphere so when translating into the z axis no resolution would be gone or do you 'map' a different texture for the different 'steps' in depth? Ofcourse the first step combined with mipmapping should do the trick as well right ? The second step seems however more the way to if you do not have massive storage and you don't get memory problems i guess.
But really, that's basically the only thing there is to it or ? I'm just talking bare bones, not about 3d cities which are added manually anyway.
-3dnewbie
Judging by your name you'll have fun doing it (sorry had to say it). It sounds really simple, but so do a lot of things: "Oh just create a virtual machine that hooks up to your brain and lets you play games with your mind." An extreme example I know, sorry to be cynical, but I've said the 'simple' line too many times and been wrong every single time.
THIS IS NOT A PERSONAL ATTACK AND I'M TRYING VERY HARD NOT TO DESTROY YOUR DREAMS.
Anyways, you're probably 1000 times better than me, go ahead and find out how complex it is, then you can tell me. Don't ask me how you would go about it, just thought I would put in my 10 cents worth.
THIS IS NOT A PERSONAL ATTACK AND I'M TRYING VERY HARD NOT TO DESTROY YOUR DREAMS.
Anyways, you're probably 1000 times better than me, go ahead and find out how complex it is, then you can tell me. Don't ask me how you would go about it, just thought I would put in my 10 cents worth.
Quote:Original post by 3dnewbie
I wonder how difficult it would be to create something like that. I mean, that basics seem to me that it wouldn't be hard at all.
1. You create a sphere, add some kb and mouse control
then comes the important question. Do you texture map one giant terabyte texture to the sphere so when translating into the z axis no resolution would be gone or do you 'map' a different texture for the different 'steps' in depth? Ofcourse the first step combined with mipmapping should do the trick as well right ? The second step seems however more the way to if you do not have massive storage and you don't get memory problems i guess.
But really, that's basically the only thing there is to it or ? I'm just talking bare bones, not about 3d cities which are added manually anyway.
-3dnewbie
As the above poster mentioned this is not an easy undertaking. If you're serious about it take a look at OpenSceneGraph.
Author
Quote:THIS IS NOT A PERSONAL ATTACK AND I'M TRYING VERY HARD NOT TO DESTROY YOUR DREAMS.
lol. had one to many newbies taking it personal eh? :-)
it's not my personal dream ( i have other things to do right now :-)) but it just seemed like such a simple thing to create. ofcourse i'm probably hugely mistaken, as you all already pointed out but i'm new to opengl and after reading up on to it a bit this just seemed not to hard.
i would like a technical answer to my first post though, that would be appreciated.
thx guys [or girls]
You'd want to load only those textures that are actually seen, rather than downloading a terabyte of pixels and hogging the systems memory. It looks like Google Earth is doing just that: move around a little and you'll have to wait for the new parts to be downloaded.
Then just flush those parts that aren't viewed anymore, perhaps based on the distance to the current view and the time that has passed since viewing it. A prediction system could be usefull here, or it could be overkill. Might be interesting to research nonetheless.
Then just flush those parts that aren't viewed anymore, perhaps based on the distance to the current view and the time that has passed since viewing it. A prediction system could be usefull here, or it could be overkill. Might be interesting to research nonetheless.
Quote:Original post by 3dnewbie
1. You create a sphere, add some kb and mouse control
Basically, yes.
Quote:then comes the important question. Do you texture map one giant terabyte texture to the sphere so when translating into the z axis no resolution would be gone or do you 'map' a different texture for the different 'steps' in depth?
You want to split it up into sections to cull large parts of the sphere that cannot be seen, especially when zoomed in to high detail. In order the show the right level of detail, you can use a tree like structure to quickly determine which polygons and textures to render, and which to discard to free up memory. One approach would be to start out with a cube, where each side is divided using a quadtree/octree.
Quote:But really, that's basically the only thing there is to it or ? I'm just talking bare bones, not about 3d cities which are added manually anyway.
Sure, it can be done. The basic idea isn't that complex, but it takes a lot of work to be able to show enough detail fast enough.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement