Tracks Lidar Track Building Strategies - Discussion - Info Pool - Progression

[edjit]

today, im going to re do what i started again, got to a point last night where the landscape and roads were too messy so... iteration 3 hahah ! hope fully this reduces the work in blender.
here is where i got to last night

was happy with how blender was handling it, very smooth but the issue was with the trees.
in the screenshot below it shows
-redline road
-green trees
as the lidar is taken from above you get no road data where the trees are covering the road, this led me to try sculpting the road back in which became problematic, lumpy roads. this was taking way too long to iron out and the tools i was using were not cutting it.

just had a thought -
in unreal engine there are landscape tools and one of those is a tool called slope i think haha,
what it allows you to do is draw a line of a gradient with width, this would be ideal for this, would be possible to bulldozer straight through the tree areas! also the flatten tool in UE is amazing unlike blender.
do wonder if there are tools like UE for blender. will search.

 

[edjit]

think i may of found what would work gonna try this, has the slope / ramp tool !!!!!!!!!!!!!!

GitHub - blackears/blenderTerrainSculpt: Tools for terrain sculpting in Blender.

Tools for terrain sculpting in Blender. Contribute to blackears/blenderTerrainSculpt development by creating an account on GitHub.

github.com

 

[edjit]

@wimdes
hey sup, was wondering if you now how to
"To build, execute the *makeDeploy.py* script in the root of the project."
its the 1st part of how to install this addon, have been trying and no luck it looks like a tool that would be very beneficial for sculpting roads.
any thoughts would be greatly appreciated ​

 

[edjit]

@wimdes​

hey sup, was wondering if you now how to​

"To build, execute the *makeDeploy.py* script in the root of the project."​

its the 1st part of how to install this addon, have been trying and no luck it looks like a tool that would be very beneficial for sculpting roads.​

any thoughts would be greatly appreciated ​

found this... seems to good for the most part, will try is and see

 

[edjit]

sup,
had a thought -
gonna try making a flat version of the track then projecting the tracks onto the terrain that has been smoothed out.
lets see how that goes.

 

[wmialil]

imo (unless you have an extremely high accuracy 3d scan) the best way is to shrinkwrap the curve/path/spline to the terrain or lidar that you have and then set your road along this spline:

Then match up the camber and width as best you can using the tilt and width of the curve itself. (you can do further tweaks with proportional editing when youve applied the road object to the curve if you want):

And with rally tracks my new workflow is basically modelling sections of road, and then modelling a few variations of alphablended road edge, all of which attach seamlessly to each other. You can have loads of RGB masks for the road section (to add variation), and if you randomize which bits of road edge go on different bits of road, you can have loads and loads of variation using only a few textures.

The edge of the road is pretty important in rally so that's why I'm dedicating more modelling + textures to it.

 

[edjit]

• wmialil
  HOLY FSH!
  GENIUS !!
  Looks amazing !
  The colour is on point, textures are mint too!
  soo stoked to try this out !!​

 

[wimdes]

@edjit my Blender knowledge is really basic, I have more experience with the tools "before" Blender

@oigolue for merging those .laz files "lasmerge" from lastools is the most obvious one. But it quickly becomes very taxing on your system as .laz is a compressed format, when opened it it will be uncompressed in system memory - that's where the "copc" format may come in handy.

The information I had in mind to start with (rough draft):

When people see the term "laserscanned" they expect the mm-precision as when the data has been acquired from a terrestrial scan, which is a slow and expensive process. While in the context of modding the public data will always have been acquired from an aerial scan, which has a much lower precision (5cm is the best I've seen). There's also mobile scanning (manual, or vehicle based), the best precision I've seen there was 2cm (but mostly around 5-15cm!)

Then the terms "accuracy" and "precision" confusingly don't quite mean the same, as greatly explained here. The 5cm precision in aerial lidar data means the road surface will be very "noisy" and has to be smoothened.

Next important parameter is points density; the number of points per m² in the data. Sometimes just 1-2/m2, best I've seen had 25-30/m2 (but higher surely exists) Like the resolution in a picture, higher resolution offering more detail, but not necessarily meaning the picture is sharp (cfr the precision of each point)

Some information on coordinate systems

Some information on the most common file formats las/laz/copc, tiles, points classification
Point cloud <> DTM <> DSM

 

[wmialil]

The colour is on point, textures are mint too!
soo stoked to try this out !!​

Colours need a bit of tweaking still actually and it's just a test track, but will definitely be using this technique in future yeah

 

[oigolue]

@edjit my Blender knowledge is really basic, I have more experience with the tools "before" Blender

@oigolue for merging those .laz files "lasmerge" from lastools is the most obvious one. But it quickly becomes very taxing on your system as .laz is a compressed format, when opened it it will be uncompressed in system memory - that's where the "copc" format may come in handy.

The information I had in mind to start with (rough draft):

When people see the term "laserscanned" they expect the mm-precision as when the data has been acquired from a terrestrial scan, which is a slow and expensive process. While in the context of modding the public data will always have been acquired from an aerial scan, which has a much lower precision (5cm is the best I've seen). There's also mobile scanning (manual, or vehicle based), the best precision I've seen there was 2cm (but mostly around 5-15cm!)

Then the terms "accuracy" and "precision" confusingly don't quite mean the same, as greatly explained here. The 5cm precision in aerial lidar data means the road surface will be very "noisy" and has to be smoothened.

Next important parameter is points density; the number of points per m² in the data. Sometimes just 1-2/m2, best I've seen had 25-30/m2 (but higher surely exists) Like the resolution in a picture, higher resolution offering more detail, but not necessarily meaning the picture is sharp (cfr the precision of each point)

Some information on coordinate systems

Some information on the most common file formats las/laz/copc, tiles, points classification
Point cloud <> DTM <> DSM

Very useful info, i'll check the .copc format to see if i can handle it, if not i may change the stage to one of the Rally de Ourense, as i found good DTM data for galicia and i prefer not to use just google elevation data.

 

[wimdes]

.copc is only for visualisation afaik, QQgis wil automatically convert laz/las to copc when you add a point cloud layer.
On the "centrodedescargas.cnig.es" site there's a DTM => MDT02 with a 2m resolution?

For lasmerge you don't need much memory, but I think for other operations down the line you'll be hindered by the 16GB of RAM again. Memory prices have come down quite a bit in the last months, may be worth looking into that anyway.

If you have the las files + a gpx of the route I can have a look to create a mesh (full detail for road + next to it, then decrease)

 

[oigolue]

.copc is only for visualisation afaik, QQgis wil automatically convert laz/las to copc when you add a point cloud layer.
On the "centrodedescargas.cnig.es" site there's a DTM => MDT02 with a 2m resolution?

For lasmerge you don't need much memory, but I think for other operations down the line you'll be hindered by the 16GB of RAM again. Memory prices have come down quite a bit in the last months, may be worth looking into that anyway.

If you have the las files + a gpx of the route I can have a look to create a mesh (full detail for road + next to it, then decrease)

Yeah i have both! That would be awesome if you don't mind, thanks

 

[oigolue]

And i'll look for more RAM, my pc can take up to 32, i don't have a lot of money rn but the sacrifice will be worth hehe

 

[wimdes]

@wmialil I don't know how you create terrain mesh starting from the las files?

What I have in mind and individual steps tested, and think it's possible to fully automate this:
Split the point cloud in a number of zones according to distance from the road, apply more thinning the further away
For instance road +2m = full detail, then next 30m = 1 point/m², next 150m = 1point/4m², further away = 1 point/25m²
Merge the zones into 1 file again and create the mesh

Would that be useful?

 

[oigolue]

@wmialil I don't know how you create terrain mesh starting from the las files?

What I have in mind and individual steps tested, and think it's possible to fully automate this:
Split the point cloud in a number of zones according to distance from the road, apply more thinning the further away
For instance road +2m = full detail, then next 30m = 1 point/m², next 150m = 1point/4m², further away = 1 point/25m²
Merge the zones into 1 file again and create the mesh

Would that be useful?

You can do it with qgis, i tried it but my problem comes just with the mesh, my pc can't handle it anymore

Take a look at this: https://assettocorsamods.net/threads/lidar-point-cloud-to-mesh-tutorial.422/

 

[wimdes]

You can do it with qgis, i tried it but my problem comes just with the mesh, my pc can't handle it anymore

Take a look at this: https://assettocorsamods.net/threads/lidar-point-cloud-to-mesh-tutorial.422/

There it's all or nothing, with the huge environments we want for rally tracks you'd need a monster pc indeed to handle the files. What I'm gunning for is keeping the maximum point cloud resolution for the road (+ just besides), then gradually decrease resolution the further away from the road. I think it should work on any pc, but the more memory you have the faster it will be finished

 

[oigolue]

There it's all or nothing, with the huge environments we want for rally tracks you'd need a monster pc indeed to handle the files. What I'm gunning for is keeping the maximum point cloud resolution for the road (+ just besides), then gradually decrease resolution the further away from the road. I think it should work on any pc, but the more memory you have the faster it will be finished

Yeah and with my pc i feel like it can't handle it but it's not like it's dying, a more optimized file would work 100%, so your idea is awesome

 

[wmialil]

@wmialil I don't know how you create terrain mesh starting from the las files?

What I have in mind and individual steps tested, and think it's possible to fully automate this:
Split the point cloud in a number of zones according to distance from the road, apply more thinning the further away
For instance road +2m = full detail, then next 30m = 1 point/m², next 150m = 1point/4m², further away = 1 point/25m²
Merge the zones into 1 file again and create the mesh

Would that be useful?

I think your method is good, I just reduce the number of points in cloudcompare and only really keep the road area + around 5m to each side.

Usually line this up with a lower detail heightmap, then shrinkwrap terrain and road meshes over the top (you never actually use the LIDAR mesh in-game in the end)

 

[wmialil]

Yeah and with my pc i feel like it can't handle it but it's not like it's dying, a more optimized file would work 100%, so your idea is awesome

You've gotta do it in small chunks if you're on an older PC/don't have a lot of RAM

 

[oigolue]

Si

I think your method is good, I just reduce the number of points in cloudcompare and only really keep the road area + around 5m to each side.

Usually line this up with a lower detail heightmap, then shrinkwrap terrain and road meshes over the top (you never actually use the LIDAR mesh in-game in the end)

Could i merge the road that i had from LiDAR data with a DTM heightmap? Thanks for the help btw!