I recently got to use a Z+F 5010C to scan our photogrammetry lab. While playing around with the Point clouds, I wondered, what Level of Detail of the Point cloud would be needed to model a laserscanned race track, like Rockingham or any iR track? And are less expensive alternatives, such as UAV photogrammetry a real alternativ to the Task? What are the relevant demands on the programmer's end and what are the racer's demands?
What demands must a point cloud meet to model a race track from it?
(13 posts, started )
What demands must a point cloud meet to model a race track from it?
As of the points density, you might get some rough idea from screenshots
Thanks. It gets me a rough idea in a way. It afterall happend 7 years ago, so it might very well be true, these screenshots show every measured point (well increased in size by some margin).
What would be interesting in this context though: What are the smallest differences the in-game track can represent. A laser scanner may be accurate between 1 cm and 0,2 mm in its spatial relation, depending on various factors. It is quite crutial to know what is the maximum there to adjust the project strategy and everything.
I took a quick look at the surveying company's Facebook page and they have many scanners of Faro. Assuming they used them back in time as well, I might get some statistical data. Maybe this gives me further clues on the topic.
What would be interesting in this context though: What are the smallest differences the in-game track can represent. A laser scanner may be accurate between 1 cm and 0,2 mm in its spatial relation, depending on various factors. It is quite crutial to know what is the maximum there to adjust the project strategy and everything.
I took a quick look at the surveying company's Facebook page and they have many scanners of Faro. Assuming they used them back in time as well, I might get some statistical data. Maybe this gives me further clues on the topic.
The company that did the scanning for us, APR Services, are flexible on how they produce the data. Their programmer was able to export the data in a way that suited us.
We asked for our point cloud data in a simple format, millions of points with position (x, y, z) and colour (r, g, b).
The laser scanner doesn't actually get colour itself but high resolution photos are taken from the same location as the scanner and this can be compared with direction to provide a colour value for the 3d points.
The scanner produces much more data nearby due to the nature of the scanning process, so the data needs to be thinned out. Thinning the data is also useful for download speed. I thinned the data further so that it was good for real time use in our editor. The point cloud can be switched on and off in our editor.
In our system there is no automatic transfer from the point cloud to actual game data. It is only a guide. It is accurate enough to ensure that objects are positioned and sized correctly within a few cm and the road surface is correct within a few mm. It allowed us to create all the bumps in the area by ensuring the track surface passes through the points in the point cloud.
So in our case there are no very strict requirements and specifications for the point cloud. We needed to make sure there were a lot of points on all track surfaces and sufficient density to get the scenery objects in the right place and the right size. The scanner was moved to various positions around the track area, and the point clouds were merged later (by APR). The merging was made accurate by placing posts in the ground that could be seen by each scanner.
For our purposes I thinned the data down to 100mm between points. It is less dense where the points were further from the scanner. The colour image can be drawn with 100mm squares, or a special accurate mode can be used where the squares are drawn with 20mm size, to get the road surface correct within a few mm by aiming for the centre of these small squares.
We asked for our point cloud data in a simple format, millions of points with position (x, y, z) and colour (r, g, b).
The laser scanner doesn't actually get colour itself but high resolution photos are taken from the same location as the scanner and this can be compared with direction to provide a colour value for the 3d points.
The scanner produces much more data nearby due to the nature of the scanning process, so the data needs to be thinned out. Thinning the data is also useful for download speed. I thinned the data further so that it was good for real time use in our editor. The point cloud can be switched on and off in our editor.
In our system there is no automatic transfer from the point cloud to actual game data. It is only a guide. It is accurate enough to ensure that objects are positioned and sized correctly within a few cm and the road surface is correct within a few mm. It allowed us to create all the bumps in the area by ensuring the track surface passes through the points in the point cloud.
So in our case there are no very strict requirements and specifications for the point cloud. We needed to make sure there were a lot of points on all track surfaces and sufficient density to get the scenery objects in the right place and the right size. The scanner was moved to various positions around the track area, and the point clouds were merged later (by APR). The merging was made accurate by placing posts in the ground that could be seen by each scanner.
For our purposes I thinned the data down to 100mm between points. It is less dense where the points were further from the scanner. The colour image can be drawn with 100mm squares, or a special accurate mode can be used where the squares are drawn with 20mm size, to get the road surface correct within a few mm by aiming for the centre of these small squares.
This stuff should be put in some sort of dev-blog. Interesting stuff, thanks Scawen.
^
Exactly, I miss all the interesting stuff the devs talk about because they're just random comments in random posts.
Exactly, I miss all the interesting stuff the devs talk about because they're just random comments in random posts.
Rockingham, recreated in Live for Speed (Mon 21 Sep 2009)
Sure, but that's not really a blog is it. And there are lots of other really nice posts that Scawen makes/has made in the past that could do with similar treatment to that.
Edit: Alright, it's the right idea, but I like more technical posts too similar to the one Scawen made in this thread. The 'Rockingham recreated in Live for Speed' page wasn't very technical and was written for a general audience.
Well, it's something. I'm also very interested in Scawen's progress so I used to check his posts. Some time ago I instead wrote a scraper to collect his, Victors (and Erics) posts that generates an RSS feed that I've subscribed my smartphone to.
The feed can be found here, and here's some more info about the scraper.
Edit: Sorry for being OT.
Not a bad solution, thanks for posting. I still reserve that separate posts with their own clear context would be nice too without occasionally having to read up through threads to work that out. 
I think we/I dragged this thread far enough off-topic.
Scawen/Eric, I'd be interested to know just how long the build process actually took in terms of man hours needed to convert the raw point data into the finished article, for instance how much of the conversion process were you guys able to automate? Too many questions really, but this is too interesting not to ask.
I think we/I dragged this thread far enough off-topic.
Scawen/Eric, I'd be interested to know just how long the build process actually took in terms of man hours needed to convert the raw point data into the finished article, for instance how much of the conversion process were you guys able to automate? Too many questions really, but this is too interesting not to ask.
Scawen,
I intend to setup a walk in facility to run 8 to 24 multiplayer simulators, using LFS commercial license.
I may be setting up near a local race track, it is small, but it will be active. It would have 3 possibly layouts, and an autocross area in center. They have a bunch of local service drives to private garage/condo units, that I would like to allow people to drive down as well. They already have cad models. Most of the surrounding facility is still under construction. On the off chance they ask me what it would cost to get their track into the simulator... How much cash/time/labor would be needed to do so?
Track is in the USA.
I intend to setup a walk in facility to run 8 to 24 multiplayer simulators, using LFS commercial license.
I may be setting up near a local race track, it is small, but it will be active. It would have 3 possibly layouts, and an autocross area in center. They have a bunch of local service drives to private garage/condo units, that I would like to allow people to drive down as well. They already have cad models. Most of the surrounding facility is still under construction. On the off chance they ask me what it would cost to get their track into the simulator... How much cash/time/labor would be needed to do so?
Track is in the USA.
Unfortunately it's not possible for us to create tracks on request at this time, even for large sums of money!
The thing is, our work for now must be towards adding new content for the S3 license and also improvements and updates to what we have.
In the future, when we have got to a certain point, I hope that LFS can be opened up to allow tracks and cars to be created by other people. It's a really big subject though and not one to get into now.
So the short answer is, you have a good question but we are fully occupied with existing projects and LFS isn't open for other content creators yet.
The thing is, our work for now must be towards adding new content for the S3 license and also improvements and updates to what we have.
In the future, when we have got to a certain point, I hope that LFS can be opened up to allow tracks and cars to be created by other people. It's a really big subject though and not one to get into now.
So the short answer is, you have a good question but we are fully occupied with existing projects and LFS isn't open for other content creators yet.
fingers crossed for new S3 content soon
traaaaaacks...
traaaaaacks...
What demands must a point cloud meet to model a race track from it?
(13 posts, started )
FGED GREDG RDFGDR GSFDG