Remember that company that promised unlimited detail for games? Well they've moved on to a slightly different market. Spatial Data sets.
Recently Euclideon have setup a demo project for Logan City Council with all of its ALS (Airborne Laser Scanning) and LiDAR (Light Detection And Ranging) points... Just as idea of the magnitude of points, the complete ALS dataset we have averages 8-10 points per square meter. That's well in excess of 100 Million points. We also have some other point cloud data (Mapped by a LiDAR Scanner that sits on the back of a Ute Like this.) Our project data set is only ~13GB created from ~50GB of ALS data and their site claims they can handle up to 140TBs.
The software flies through the data very smoothly on a mid-spec machine (It's 'High-Spec' at work, but my 2 yo gaming computer is better than it) and it doesn't skip a beat or lurch or anything.
We have 3 different demo projects. 1 with Aerial Photography, 1 with colourised points and 1 with Aerial and the 3D street scan + photos (Think Google Street view, but with higher res and you can actually measure off everything)
The software only has basic measuring tools at the moment (something they are going to build on) but can do Distance/Height etc. Controls took me a few min to get use to, and some others are pretty funny to watch as they try to control what handles like a spectator control in a FPS. Most people get the hang of it fairly quickly though. Another thing we also have in most LiDAR data is intensity information (how strong the return beam was) which is another thing they're working towards including in their software. This data can be used to map how thick vegetation and even what type of tree the Light point hit etc.
I didn't get a chance to have a sit down with the Devs (Boss went along) so I couldn't get many technical questions out. But I will get a chance to ask questions (one of which will be if they still have any focus on Games development) so if you have any to add I'll be happy to pass them along.
All in all it's a pretty cool package and I'm more excited for this stuff than I am for their gaming advancements. I'm very keen to see just how far they can take it. If we could combine this stuff with some of the flood mapping and prediction software it would be insanely cool. I'd love to watch exactly how the entire city would flood, real time, with programmed rain events.
Shopping Centre Entrance, take a look at the 1/2 cars for how much detail was captured.
X,Y,Z points on the Shopping Centre Sign. All have the photo overlayed on them.
Point Detail on some Overhead Power Lines
Little Boxes on the hill side and they all look just like Ticky-Tacky and they all look just the same.
Colurised ALS data.
Scary, that raises some massive issues about privacy.
my scrollwheel just moves the whole thread
I assume there is meant to be interactivity that explains why this is good? Because right now all I'm seeing is some very s*** 2D images with the newspaper filter on them.
last edited by thermite at 10:46:28 10/Dec/12
Yeah, the fact that you can display and fly through literally MILLIONS of points of data without any loading, lurching, jumping, rendering times.
my scrollwheel just moves the whole thread
you seriously didn't realise that its just screenshots from the software?
Looks pretty cool though.
Looks like awesome tech.
Imagine something like google street view... instead of being 'imagery' a rendered model.
Do they need investors (looks like a good option for the geospatial tech. end of the world/market)?
'm very keen to see just how far they can take it. If we could combine this stuff with some of the flood mapping and prediction software it would be insanely cool. I'd love to watch exactly how the entire city would flood, real time, with programmed rain events.
i rekn! that'd be swoit! :D
this could be a cool way to make maps in games.. imagine rolling through logan hyperdome FPS style.
Does this triangulate data? Can you get volumes from it? Otherwise I don't see the point besides showing stuff off visually
The company has been around for a while, as far as tech timelines go, but it's probably only been ~2 years that they have been working on spatial data. Not really 'ages'
At the moment DK it only has rudementry measurig tools, so no triangulation/volumes. Showing off stuff visually is one of it's main points though. There are so many people out there that could use spatial data like this that simply cannot see in 3D from a plan. I can look at a contour map and know how steep those gullies are and how tall that mountain is but many people cant.
Another use for data like this (though this was already possible in other packages, this makes it more user friendly... and a hell of a lot smoother) is for things like finding how much clearence Power Lines have from trees. Energex could do a low pass on their power lines and find out exactally which/how many trees need to be cut/trimmed and which can wait another year or 2. Private sector is starting to use Drones now as well so the collection of this type of data has and will become a lot cheaper.
Thanks a lot for this post. This is very interesting in my field of intelligent vehicle systems. We work with LIDAR all the time but I did not know of these companies and councils collecting this kind of data.
It's great to find someone who has had a hands-on with this tech - thanks Scooter.
Do you have any idea what the FPS is? What is the loading time like?
I'm more interested in this for gaming purposes but can see how it would be useful for a variety of purposes. Obviously for games it won't be acceptable to have gaps inbetween the points like from the lidar data - I wonder how it will handle denser scenes.
Wouldn't be able to tell FPS but I can fly through it at a pretty impressive speed without it skipping or jumping. Loading time is <5 seconds from the desktop to an open project database.
The index is an octree, hence the ".OCT3" extension. Compression for Mr. Dell means the "recycling of common structures" i.e., storing the locational codes' common prefixes and common suffixes just once. Recycling common prefixes gives octrees (because locational codes in 3-D are ogdoadic sequences). Recycling common suffixes gives DAGs (because hardly do non-leaf nodes store anything except the indices of their children. A leaf's color is likely stored in lieu of its index. Already this, with 32-bit indices, results in about 2:7 compression).
Any news? An interesting question: was Bruce Dell a Christian when the octants' rejection method occurred to him?
I'd like to know if each of these voxel points can store light, specular, opacity or any other sort of information, I assume it's currently storing only translation and colour (RGB) values. I'm guessing if you start adding more info per point the data size is going to increase exponentially.