Outerra forum
Outerra Engine => Technology => Topic started by: Redrobes on August 04, 2011, 07:49:46 pm
-
Ravs off of the Cartographers guild (dot com for all those not familiar...) has posted an interesting page about terrain scanning and 3D terrain vis.
Its a Swedish company processing data from either Saab or a Saab based technology outfit. Check it out.
http://www.youtube.com/watch?v=gSmunh6NIQI
http://www.youtube.com/watch?v=BaahKhvO_E4&NR
http://www.youtube.com/watch?v=CNemPTHOKWg
-
i hate tech like this not being used in a flight simulator :( :mad:
-
Nokia maps is using this technology, they already scanned a few entire cities and their surroundings. Here you can appreciate the level of detail of the moutains next to Cape Town : http://maps.nokia.com/3D/#|-33.95580122|18.413001786|13.92|76.67|69.58|0|26| (it will ask you to install a little plugin)
Indeed it would be perfect for a flight simulator, but it's not very flexible if you need more interactivity (lighting, physics...)
-
That's some crazy stuff. Wonder how much it costs to sample a city like that. Probably alot.
-
how is this possible??
how do they know the exact shape of each house, and where the heck do the textures come from?
i can understand googlemaps 3d buildings as they are handmade, but i cant understand this...
-
They fly a plane over the terrain in some kind of raster flight path. The plane has a number of camera on board. The number was said to be secret but I could believe it to be a lot - like 16 or something like that. They have been bolted down with fairly accurate position and such that they overlap. They did say they were digital SLR cameras.
So they know the planes position very accurately - probably to about a few cm's. They know the flight path at the time to a similar accuracy and they have the camera positions and directions known. Then they match up multiple photos from different shots at different times. Because you know where you were at the two times you can see the differences that show up on the spacing between features across two or more shots. This gives you Z depth. Working back from that plus the planes position can give the XYZ position of the ground terrain. Normally you still get shadow zones from the parts of the photos which were obscured but if your taking many many shots as part of a raster flight path then there is going to be very little left obscured. Basically only stuff underneath a cover.
You can do some of the math and processing to scan in a *sparse* point cloud easily using your own digital cam and some free software. I have wrote a blog about that - you might need a login to see pics, not sure:
http://www.cartographersguild.com/blog.php?1311-Redrobes-blog-about-playing-with-the-technical-side-of-graphics.
These guys tho are getting more than a sparse point cloud out of it and are post processing it a bit more to generate the buildings as vectored meshes. I can see from the video that they go beyond this to work out which meshes are buildings and which are roads or terrain etc. They are classifying the mesh output as well. So they are well beyond the normal consumer state. But then they are military and well funded...
I didn't realize Nokia have them. That's quite cool. All the demos were on iPads :o
-
thanks for the explanation, very interesting stuff!
-
Damn ... it breaks my heart to see such lovely aireal 3d images and to know that its NOT available for Flight simmers.
-
Its a Swedish company
It must be, look at the title only..
Oslo, Norway not Sweden
as of course Oslo resits in Norway :P
KUDOS to the author :)
-
Its a Swedish company
It must be, look at the title only..
Oslo, Norway not Sweden
as of course Oslo resits in Norway :P
KUDOS to the author :)
No, it's a Swedish company, i don't get how you made the connection "They showcased it with Oslo, that must make they Norwegian!" or what? ???
http://en.wikipedia.org/wiki/Saab (http://en.wikipedia.org/wiki/Saab)
They also make cars.
http://en.wikipedia.org/wiki/Saab_Automobile (http://en.wikipedia.org/wiki/Saab_Automobile)
Here, read this Wikipedia page and remember not to make stupid assumptions like that anymore.
-
If this would be implemented in outerra, the game I always dreamed about would became truth....
-
If this would be implemented in outerra, the game I always dreamed about would became truth....
It's very expensive, so i don't really think outerra will use it.
To be honest, it's probably cheaper to manually model and texture everything than using this method.
-
I saw it was very expensive, but I wanted to say my opinion.
-
Hey you guys should take a peek at the new Nokia 3D maps using webGL. Its cool and try flying around NY Manhattan.
http://maps3d.svc.nokia.com/webgl/index.html (http://maps3d.svc.nokia.com/webgl/index.html)
-
No plugin is needed anymore, that's cool, and the point of view is less limited than before, otherwise no big improvement, except internally maybe.
-
Its a Swedish company
It must be, look at the title only..
Oslo, Norway not Sweden
as of course Oslo resits in Norway :P
KUDOS to the author :)
No, it's a Swedish company, i don't get how you made the connection "They showcased it with Oslo, that must make they Norwegian!" or what? ???
Oslo is the capital city of Norway. Yes, SAAB is from Sweden, our neighboring country. SAAB just took a visit to Oslo to test their technology.
Regarding using the technology: Use it to a certain degree, which Outerra is doing. By that I mean that saving a whole world that detailed on computer today is too heavy. It takes too much space, meaning it's heavy to load too. What is often done to save the problem is making it procedural, preprogrammed. What I think Outerra is doing is taking a simple mesh of the earth, then making a system for details it with general presets.
-
Just for your info, this techno has been purchased by Apple months ago, Apple bought also another Canadian company who was doing one cross-browser/platform globe engine ;)
In fact what is behind is that Apple is preparing one new app to compete with GoogleEarth.
As for data quantity of this kind, for the entire Earth, it is only a question of Teras or Petas, and of cloud computing/streaming, and I do not think that these poses much problem to "monters" like Apple.
-
Again, the big 3 , Microsoft, Google and Apple are in big maneuvres concerning this 3D cities subject... MS and Apple are acquiring technologies, so far Googl has counted a lot on people working freely for them with Sketchup (but they have sold Sketchup to another company)... so we will see new things soon.
Coming back to Saab's spin off, it is not so innovative, but of course they have a rapid acquiring system and surely a nice photogrammetric process, and they are able to "scan" large areas quite easily. But there are limitations and it is as always a question of approximation.
Let's zoom close to the Empire State building in NY and there you will see the level of detail of their meshes and textures.
Also you cannot lower the altitude, your wiew position is always close to the one they used when they took the pictures, and their good realism trick is here : you view as built (or as acquired). So navigation, viewing limits!
do not think to go at street level or at low altitude with such models. And it is fully explicit datas, nothing procedural or smart.
Even the trees are pretty " blocky ". Think that there is so far no realistic lighting, no possible behaviors on these datas
( or I miss something?). I'd like to see them render NY at night, maybe they will vectorize their cities in the future with a classification (windows, trees, light poles, cars)? I wonder...
(http://www.cadasys.com/images/NY-Empire-State-Building.jpg)
Some other companies and technos start to do smarter, mixing vector datas and aerial ground imagery, with points for insertion of real 3D trees, 3D lightpoles, procedural roads or streets, classes on the objects, with behaviors and so on... and with better buildings, more accurate, with hi-res texturation.
The future is there according to me : a mix of procedural and explicit.
I dream of realistic but intelligent datas, where you can take off, from the CITY street level to space : it is what Outerra is able to do and I think it is only a question of doing it the best way.
Dear Cameni and Angrypig, could you please give us your thoughts about these "civilization" non natural environments?
is there any fractalization possible in such man made places ? is for you, CityGML the best model ?
-
Depends on what you want to achieve: maximum realism or maximum believability. The former needs huge amounts of data combined with clever techniques to acquire and render them, and that's where Google/MS/Apple is heading, for use in navigation and maps and such. The latter can do with much less real data but augmented by probabilistic procedural models to make things more immersive, to make you believe a world like that is possible and feels real, with primary use in games, simulators and movies.
Obviously it's not black & white; the procedural approach of OT already combines real data with algorithms that refine them, as opposed to a purely procedural approach of most of other procedural engines.
These two can be combined; making 3D models from captured data is already a form of vectorization, but it needs to get to a different level to be usable in the procedural model. But I think it all iterates towards that - in the future there will be a tool that's capable to extract not only a 3D contour model, but also the pattern, a procedural equation, outputting metadata with procedural definition of how the building was assembled. And that's because the buildings are highly regular, they were procedurally built in the first place.
We are approaching things from the game/sim point of view, and from the procedural side. We aim for procedurally generated buildings and cities, while the generator parameters could be also provided externally. You could have cities that were completely generated, or you could create the basic roads and define city blocks, assign block types and have the generator fill it with a finer road network and buildings pertaining to given block type, or define individual parameters for building(s) when you want a specific one at a specific place.
-
Thank you for giving us your feeling about this. Yes it cannot be black or white, and you are right saying that it truly depends on the final usage.
For games and movies, I can assume that it must be beleivable, but not necessarily nearly exact reproduction of real places.
And of course it is much easier to create fake or Sci-Fi cities (non existing urban places),
But if the game, scenes are in real places, there is no other chance than representing these real places.
A game action close to the Tour Eiffel must have the Tour and the Paris quarter around.
This is the same with most sim users (real simmers or mil simmers) who want to see their environment as close as the real one.
Flight simmers will want to fly or hower above New York or Paris or Bratislava.
See the previous comments where the guys complain about the fact that they don't have such C3 techno models in their actual flight simulators
I watched quickly the Monaco F1 grand prix this weekend, and of course I imagine that the race gamers have some pretty realistic Monaco circuit too, have they?
If we talk about GIS users, because your engine is still generic and powerful, then it is becoming another matter, because not only they look for reality but they want it very accurate, and they want to filter and make many "stupid" things.
To conclude, and as your generator does already work with external datas (terrain and others), then mixing real and procedural won't be difficult :)
-
But if the game, scenes are in real places, there is no other chance than representing these real places.
A game action close to the Tour Eiffel must have the Tour and the Paris quarter around.
Well, there's this word "representing" that allows for some uncertainty. We all know how Eiffel generally looks. But you won't bet your wife's diamonds on whether a particular rivet (http://fc03.deviantart.net/fs71/i/2010/219/2/f/Eiffel_Tower_close_up_by_Kyanelia.jpg) is or isn't there in reality. What buildings are around? And a bit further from that? For a game taking place in Paris the rendition of the city must be only as good as people's image what Paris looks is. Of course, it's different for the citizens vs. me, just as the Eiffel is different for the painter doing the maintenance vs. the tourist :)
Just as lossy image compression is acceptable, a lossy geometrical and conceptual compression would be acceptable as well, and the compression ratio acceptability threshold depends on the target group. Simmers need to have certain information-rich regions more realistically rendered, but they don't need to see every common building looking exactly as in RL. If I asked them to choose between a completely real but blurry or deformed (because of amount of data limits), statically lighted rendering and a procedural one where not every building looks exactly like it does in RL but where the city looks real and believable ... what would they pick?
-
Sure they would choose the last one, the city which looks real and believable,
but creating a procedural believable Paris or London is anyway not so trivial :-\
Of course you can calculate some probabilities, but you cannot prevent that some of the procedural objects be generated, like you said above, with "input" of minimum RL external datas, these datas coming from various sources and formats.
Like GIS (shapefile or others), for the streets, the buildings footprints, poles, bridges, etc,
a fine DTM or DSM model, maybe some hi-res SAT imagery on some places.
Well nothing is impossible, but what I think, is that the problem is more on "getting these datas for free, or for a low price" : this might not be so easy...
As for the RIVETS, if we want to have such a level of detail/realism, then I have no doubt about the fact that you are skilled enough to apply one of your clever law / algorithm, with some probability, perturbation, propagation functions... One who can render probable moving grasses, should have NO difficulty in Rivets ;D ;D :D
BTW you could put RIVETS on your to-do list ;D because at a moment you will have solved all that OT is still missing 8) what else do you still have to do? volumetric clouds, winds, geology, biomes, rivers, lakes, night, HDR, importer, physics, well... not so much ;D ;D ;D
-
More info about the Nokia maps and the LIDAR scanning tech used to make them.
http://m.theatlantic.com/technology/archive/2012/10/the-forgotten-mapmaker-nokia-has-better-maps-than-apple-and-maybe-even-google/263150/ (http://m.theatlantic.com/technology/archive/2012/10/the-forgotten-mapmaker-nokia-has-better-maps-than-apple-and-maybe-even-google/263150/)
-
Yes, they show an interesting presentation about their LiDAR process here :
NAVTEQ 2010: NAVTEQ True, Understanding the Collection Process (http://www.youtube.com/watch?v=Z_24s5pzBQA#ws)
-
Yes, they show an interesting presentation about their LiDAR process here
Damn that's pretty cool. But I wonder how they'll deal with the roofs of the building and homes and other stuffs their cam can't reach.
Add the data from satellite/plane view ?
Also, I would not say no to a GTA game made this way :D
-
Sounds like that thing needs to lap the Nurburgring a few times.. And hand Cameni the data.
-
to XZS : for the " above ground " features, they can use aerial LIDAR or photos or other systems, (the military are using so many sensor technologies, some like SAR synthetic apperture radar, wich are able to "scan" through clouds or fog...)
In fact, if you start to mix all these datas coming from various existing sensors and techs, you solve the problem.
What is much more tricky is to merge and rend all these datas user-friendly wich means : clean, simplified, non redundant, light, etc...
According to me, this is only feasible with skilled human operators who post process them.
Add the costs of acquisition + costs of smart post process, and then comes the big question : having clean 3D datas from cities or from any other part of the Planet, Okay, but for what usage ?
-
well as mentioned in the video you could turn any maps application into a second advertisement place.
-
In fact there are many many applications but the costs of acquisition of these 3D datas are very very expensive and apart of the military, who owns the hardware equipments powerful enough to visualize such quantities of 3D stuff in real time ?
-
well i would think you could do it similar to outerra with a lot of lod-ing and modern hardware shouldn't be such a problem what could be a problem is the bandwidth necessary to transmit all that data
-
yes I do agree you Chaoz, would be interesting to test in Outerra with a large 3D city model...