Tracks Bathurst LiDAR reboot

As an experiment, I'm going to kick off an open-source track project. Anyone who wants to contribute is welcome to!

The idea here is that future sims/mods/etc will not be in some grey area about using and converting the track. The plan is to release everything under a Creative Commons licence, blender files and all. Also that way if I get hit by a bus or whatever the data can survive.

So, I got started last night in QGIS. Got 4x4km of the 1m LiDAR DEM into Blender and draped 20cm/pixel ortho-photos over it. The scene is properly geo-referenced using the GDA94 zone 55 (EPSG:28355) projection at an origin of 737420, 6295420 in that co-ordinate reference system.

This should help to avoid the mercator scale issue no-one seems to bother engineering out of most tracks.. FWIW, the only Bathurst I've driven which I think got this right is in Gran Turismo. Image to illustrate difference between Mercator map and actual country geometry:


Was a bit ambitious with the DEM though, 16 million vertices in one object, going to divide and conquer it later today.
unknown.png

unknown.png

unknown.png
 
Thanks GT VIRUS! that'll come in handy!

I've cropped the 1m DEM down to 1550x2300 meters and now I can press tab (to enter vertex edit mode) without Blender exploding! :D
unknown.png

unknown.png


The photos seem to be free for public use, with the proviso of including a Copyright notice. There's 2m, 5m and 30m DEMs available under Creative Commons for the rest of the visual mesh! On to the background terrain next!
 
One handy thing you can do is duplicate the mesh, apply a Decimate modifier to it (default collapse with a ratio of about 0.1 is probably the best option) and then only leave that duplicate visible most of the time, makes it much lighter on blender. Use the original DEM whenever you're directly referring to an object (eg. for shrinkwrap target height)
 
Good idea Stereo, thanks, what I did on my other DEM/DSMs was use Blender's limited dissolve with a low angle to preserve as much geometry as possible, result from that is collapsing it into n-gons (can be done in rapidlasso too I think). A lot of redundant vertices left-over from the raster import which have to go either way :)
 
What we have after one day. Or less. Heatwave here atm and had to put out a fire in a paddock twice while trying to do this...

Screenshot_ks_toyota_ae86_tuned_mount_panorama_circuit_18-2-119-21-29-40.png

Screenshot_ks_toyota_ae86_tuned_mount_panorama_circuit_18-2-119-21-29-12.png

Screenshot_ks_toyota_ae86_tuned_mount_panorama_circuit_18-2-119-21-30-1.png

Screenshot_ks_toyota_ae86_tuned_mount_panorama_circuit_18-2-119-21-28-46.png
Very exciting project good luck with it, was such a pity with the last one that was WIP. :)
 
Thanks CC! Yeah that one looked good, I'm not the greatest Blender pilot or anything but I'm giving it a go

Here's what driving on raw LiDAR looks like:
Wow thats really nice that data, I done 1000,s of laps on this track, was kinda my speciality at one point,
one of the most challenging tracks I think on the world, to carry the speed and not hit anything is quite tricky, I love it,

looks so rough this data will take a lot of smoothing out in places but the track surface and elevation changes looks spot on,

what I would do in this situation to be quite honest with you,
is find a car in Forza what AC has, run 1 lap in forza,
and run the same laps in AC with the same car, but in AC match the forza cameras mm perfect same FOV same everything,
then make a side by side comparison and see how far away it is, driving the same lines ect,
although I do feel youll need a better surface than that one, so maybe run a slower car :)
 
This would be so awesome! Thanks for doing another approach to Bathurst!
Driving the ac and rF2 versions and then looking at a real life onboard it looks pretty well but some parts just don't really match. A bit of input differences and the driving just becomes a little different to the real thing sadly.

Your project already looks great! Hopefully some experienced guys will help you out! I sadly have absolutely no clue about modding...

One question already: since you want to do this open sourced, maybe you could try to get some rF2 guys in or AMS or whatever so this more accurate version might see the light in other Sims too without the usual copyright/rip off/conversion problems :)

I wish you great success with this, lots of fun and to be hours of being stuck (it always happen at some point, hehe) not become too devastating :inlove::thumbsup:
 
Wow thats really nice that data, I done 1000,s of laps on this track, was kinda my speciality at one point,
one of the most challenging tracks I think on the world, to carry the speed and not hit anything is quite tricky, I love it,

looks so rough this data will take a lot of smoothing out in places but the track surface and elevation changes looks spot on,

what I would do in this situation to be quite honest with you,
is find a car in Forza what AC has, run 1 lap in forza,
and run the same laps in AC with the same car, but in AC match the forza cameras mm perfect same FOV same everything,
then make a side by side comparison and see how far away it is, driving the same lines ect,
although I do feel youll need a better surface than that one, so maybe run a slower car :)

I have to completely disagree with you here CC, do not compare it to the forza one in general. It's crap and shouldn't be anywhere near this project
 
I have to completely disagree with you here CC, do not compare it to the forza one in general. It's crap and shouldn't be anywhere near this project
Well I was under the impression it was scanned the track, if it is scanned then obviously its accurate,
if you are saying its wrong the track itself, then it must not be scanned,
but I saw loads of features and videos and stuff when they was scanning them ect, I can recall in my mind them anyways, its the surface layout you want the same,
no game accurately portrays the landscapes and track side objects 100% perfect, they will all differ slightly, but like I say was presuming was the best resource anywhere,
if its scanned in iracing then use that I didnt realize they had one dont play it.:)
 
You are going to want the road surface to be as high res as possible to build the layout. The landscape can be much lower resolution. This usually means isolating just the road surface with the highest sampling rate possible. Then the land can be much much lower resolution. At this stage you don't even need or want the land in there as the track layout should be built first up to the guard rails. Then the land comes in later.
 
Here is an example. First is the full scan (this is a ground scan so it won't have surrounding terrain). Importing just this into blender would kill it.
Capture.JPG


So I cut out the road up to the guard rails. This was fine in blender and I was able to keep a very high res of the data.
Capture1.JPG


I would then resample the terrain to a much lower resolution

Capture2.JPG


I know you are dealing with a DEM and not a raw point cloud but you can use CC to "scan" the DEM into points where you can then do the simple resampling. That is how we worked with the Mosport data as that was also a DEM.
 
Thanks for your input LilSki!

The cut-track-out-from-DEM approach is exactly what I'm doing on the Summit Road project, just using box and circle selections, then P, S to separate as a new object.

Spent way too much time making this track outline map for the UI in QGIS hehe :whistling:
outline.png
 
Last edited:
Geoscience 'straya's SRTM-derived 30m DEM meets the LiDAR (albeit with some overlap):
mt_panorama_lidar_srtm_blend.png


Also draped it with 10m/pixel photos and translated the whole scene down 600 meters to get everything closer to the physics engine's origin. Elevation of the LiDAR DEM ranged from 694 to 878 meters.
Screenshot_ks_nissan_skyline_r34_mount_panorama_circuit_19-2-119-21-41-24.png


Started on a limited dissolve of the background DEM, 2.2 million vertices, might take a day or two.. decimating/resampling is faster but I want to try and preserve as much geometry of the terrain as possible (and this method worked for me with Jaxa's ALOS DSM already).
 
Last edited:
Still processing the dissolve :O_o: and I got some other projects on the go but made one failed attempt at shrink-wrapping a road to the DEM, 1m is still too coarse a resolution for acceptable driving, and apparently physical objects don't like being closely layered on top of one-another... oops

Anyone who wants them can get the blender files etc here:
WARNING: the track is not really drive-able and the DEM is not optimised yet, so if you get any FPS you are lucky :laugh:

will update as I progress but this is a big one, crap, what have I gotten into.. :laugh:
 

Latest News

Online or Offline racing?

  • 100% online racing

    Votes: 96 7.8%
  • 75% online 25% offline

    Votes: 130 10.5%
  • 50% online 50% offline

    Votes: 175 14.2%
  • 25% online 75% offline

    Votes: 348 28.2%
  • 100% offline racing

    Votes: 480 38.9%
  • Something else, explain in comment

    Votes: 5 0.4%
Back
Top