TOOL LIDAR data - available for free for England
- Thread starter assettocorsamods
- Start date
Pixelchaser
Well-Known Member
im going to texture that Oulton Park mesh .
edit: still a few metres out. from absolute precision with the uv. but its nice to see it with some reference.
edit: still a few metres out. from absolute precision with the uv. but its nice to see it with some reference.
Mr Whippy
Active Member
Yeah with the reference features over the top it gets better to see exactly what you're looking at.
Like you say Pixelchaser, there is still lots of work to tidy these meshes up and get what you want from them, but compared to what we had before with regards to Z data, these really help get the basic track and flows of stuff up and running nicely from the off!
Is that the highest res aerial imagery for the location Pixelchaser?
I've yet to have a play with draping imagery over the DSM, I might have a bash with some of my local roads and see how good it looks.
My main worry is stuff like shadows, feature changes, and parallax issues. I can bet lots of details at a fine level deviate a bit.
Again as you've mentioned earlier, this data isn't a panacea for track making, so we need to realise that it helps with some bits, but might actually confuse with other parts of builds!
For instance I've generated a 50cm DSM from the data for Leeds city centre. Buildings still are not defined really nicely as expected, since all their edges are 50cm leaning in at the top. That means a building could be 50cm out in X or Y vs reality, or +/- 1m too wide or narrow if building from the DSM as reference.
In those cases aerial imagery is more precise. Lots of pros and cons to everything here still!
I've even got a point cloud from aerial photogrammetry from the area and the datasets align over each other nicely, but I've still used aerial imagery high res stills for the street outlines and actual building rooftop outline bounds!
I'll be interested to see what the Environment Agency point clouds look like when they finally release them (if they eventually do any way), since they may hold quite a bit more data for our needs, vs these rationalised grids which are probably quite lossy.
I've still been looking but can't find anything on the website downloaders.
Like you say Pixelchaser, there is still lots of work to tidy these meshes up and get what you want from them, but compared to what we had before with regards to Z data, these really help get the basic track and flows of stuff up and running nicely from the off!
Is that the highest res aerial imagery for the location Pixelchaser?
I've yet to have a play with draping imagery over the DSM, I might have a bash with some of my local roads and see how good it looks.
My main worry is stuff like shadows, feature changes, and parallax issues. I can bet lots of details at a fine level deviate a bit.
Again as you've mentioned earlier, this data isn't a panacea for track making, so we need to realise that it helps with some bits, but might actually confuse with other parts of builds!
For instance I've generated a 50cm DSM from the data for Leeds city centre. Buildings still are not defined really nicely as expected, since all their edges are 50cm leaning in at the top. That means a building could be 50cm out in X or Y vs reality, or +/- 1m too wide or narrow if building from the DSM as reference.
In those cases aerial imagery is more precise. Lots of pros and cons to everything here still!
I've even got a point cloud from aerial photogrammetry from the area and the datasets align over each other nicely, but I've still used aerial imagery high res stills for the street outlines and actual building rooftop outline bounds!
I'll be interested to see what the Environment Agency point clouds look like when they finally release them (if they eventually do any way), since they may hold quite a bit more data for our needs, vs these rationalised grids which are probably quite lossy.
I've still been looking but can't find anything on the website downloaders.
Pixelchaser
Well-Known Member
nope, if I could lay a 500000 x 500000 image I would. so I settled for 16000 x 16000 converted down to 8192. zoom level 20 (maybe 21) data
its hard to line up too because the site is 2K x 1.8K. and our games need power of 2 textures (square in this instance)
Last edited:
Mr Whippy
Active Member
The way I've done these things before is grab the highest res imagery, get it aligned right, then do a planar texture to the exact dimensions of the base mesh.
Then I "render to texture" at whatever dimensions I want to test with, which obviously makes a distorted texture if the grid isn't square, but it's not so anamorphic and like you know it can be game engine friendly then for whatever purposes.
I've even done hybrid things where I have a low density planar grid with the reference imagery on it, then I've sub-d'd it, but tweak the base mesh XY vert locations so the texture can be shifted a bit to match the details on the DSM or laser scan details etc, but done smoothly via the sub-d refinement.
Then I bake that to a new texture which is 'corrected', at least to references where there is no parallax drift. Still a bit rough but quite often aerial imagery is just out and needs a forcing hand
I once had a laser scanned road which was about 5m out vs the aerial imagery, which 5km was lined up perfectly. Simply put the parallax in the aerial imagery due to terrain slopes caused the drift. I spent hours just correcting the aerial imagery to the scan data as elegantly as I could so all the stuff I then built from aerial imagery or textured baked from it lined up right
Then I "render to texture" at whatever dimensions I want to test with, which obviously makes a distorted texture if the grid isn't square, but it's not so anamorphic and like you know it can be game engine friendly then for whatever purposes.
I've even done hybrid things where I have a low density planar grid with the reference imagery on it, then I've sub-d'd it, but tweak the base mesh XY vert locations so the texture can be shifted a bit to match the details on the DSM or laser scan details etc, but done smoothly via the sub-d refinement.
Then I bake that to a new texture which is 'corrected', at least to references where there is no parallax drift. Still a bit rough but quite often aerial imagery is just out and needs a forcing hand
I once had a laser scanned road which was about 5m out vs the aerial imagery, which 5km was lined up perfectly. Simply put the parallax in the aerial imagery due to terrain slopes caused the drift. I spent hours just correcting the aerial imagery to the scan data as elegantly as I could so all the stuff I then built from aerial imagery or textured baked from it lined up right
Pixelchaser
Well-Known Member
your on a very technical level there man, its marvelous, sadly sometimes its above my level of thinking, im mainly a texture guy.
ive just tried to lay a texture down it and it will never work unless its done another way. a stich image from SAS is producing an image with geometry upto 10 metres out on 1 side whilst a point on the other side is correctly aligned. so 6/7 hours trying, I cant do it. the real problem being my computer can not UV unwrap that 3.6 million poly object. edit:: annoying thing about oulton park is there are more markers for alignment outside the track that inside or I would delete everything outwith the confines of the circuit.
it is amazing seeing an area aligned though, everything instantly becomes easy to relate too. if we can get these reference models and texture correct then it really will be a great reference. it pops right out at you when its right and shouts "MAKE ME NOOOOWWWW !"
ive just tried to lay a texture down it and it will never work unless its done another way. a stich image from SAS is producing an image with geometry upto 10 metres out on 1 side whilst a point on the other side is correctly aligned. so 6/7 hours trying, I cant do it. the real problem being my computer can not UV unwrap that 3.6 million poly object. edit:: annoying thing about oulton park is there are more markers for alignment outside the track that inside or I would delete everything outwith the confines of the circuit.
it is amazing seeing an area aligned though, everything instantly becomes easy to relate too. if we can get these reference models and texture correct then it really will be a great reference. it pops right out at you when its right and shouts "MAKE ME NOOOOWWWW !"
Last edited:
Mr Whippy
Active Member
Haha, yes the 'make me noooowwww!' appeal of a textured DSM is hard to ignore
Yeah texturing that 3mil+ geometry is gonna be a pain.
I might have a bash at it in a few weeks.
My plan would be to possibly render out the DSM to a high res image so I could see the features, then use a distort tool in PS to slowly tweak things to match.
Or do something similar within Max.
If this area were covered with one image (possible to figure it out by checking over the whole site for discontinuities), you could in theory render out a height map (z buffer depth basically), then use a Z depth map in After Effects to pixel shift in different directions (possibly also with a falloff across the image)... that might get things lining up nicely, or more nicely.
But that ignores lots of other fine variables too.
Even on a fairly 'top down' image, a few metres can mean 4 or 5 pixels on a sharp aerial image, which is quite a lot of inaccuracy!
I'm working on a city centre track right now and there are 1m wide discontinuities between some aerial images where they've been overlapped. The building tops line up ok, but the pavement and streets have a big offset!
I think drones are gonna be better for getting imagery for tracks soon. Just fly the perimeter at a decent altitude and shoot sideways into the location.
Hmmm
Yeah texturing that 3mil+ geometry is gonna be a pain.
I might have a bash at it in a few weeks.
My plan would be to possibly render out the DSM to a high res image so I could see the features, then use a distort tool in PS to slowly tweak things to match.
Or do something similar within Max.
If this area were covered with one image (possible to figure it out by checking over the whole site for discontinuities), you could in theory render out a height map (z buffer depth basically), then use a Z depth map in After Effects to pixel shift in different directions (possibly also with a falloff across the image)... that might get things lining up nicely, or more nicely.
But that ignores lots of other fine variables too.
Even on a fairly 'top down' image, a few metres can mean 4 or 5 pixels on a sharp aerial image, which is quite a lot of inaccuracy!
I'm working on a city centre track right now and there are 1m wide discontinuities between some aerial images where they've been overlapped. The building tops line up ok, but the pavement and streets have a big offset!
I think drones are gonna be better for getting imagery for tracks soon. Just fly the perimeter at a decent altitude and shoot sideways into the location.
Hmmm
Mr Whippy
Active Member
Just FYI, the Environment Agency LIDAR (LAZ) data is now available for many parts of the country.
These files load right up in CloudCompare and then you can do what you want with them from there, like meshing, using directly in 3D apps, whatever.
Where they are available the point density seems much higher than the resolution in the grids (ie, you could probably push a 10cm grid from the data with decent sampling and get usable data)
Also there is more data on side surfaces of objects depending on the angle of the flight. Ie, some buildings you actually get side-on data. Trees look much better etc.
This is an area near where I live.
The densest DSM grid mesh data available here is 1m, but here in the LAZ file there is sufficient data to actually define objects down to around +/- 25cm, so it's WELL worth using the LAZ data as a base for anything if you can get it for the area you want to work with.
Also they're super-light in a point cloud app, vs the huge 25cm DSM grids which grind conventional 3D apps to a crawl!
Cheers
Dave
These files load right up in CloudCompare and then you can do what you want with them from there, like meshing, using directly in 3D apps, whatever.
Where they are available the point density seems much higher than the resolution in the grids (ie, you could probably push a 10cm grid from the data with decent sampling and get usable data)
Also there is more data on side surfaces of objects depending on the angle of the flight. Ie, some buildings you actually get side-on data. Trees look much better etc.
This is an area near where I live.
The densest DSM grid mesh data available here is 1m, but here in the LAZ file there is sufficient data to actually define objects down to around +/- 25cm, so it's WELL worth using the LAZ data as a base for anything if you can get it for the area you want to work with.
Also they're super-light in a point cloud app, vs the huge 25cm DSM grids which grind conventional 3D apps to a crawl!
Cheers
Dave
Last edited:
Willy Wale
Member
Not sure if the UK street furniture reference thread started or not (I did search for it) but anyway the links below may be of use
Department for transport Traffic sign manual - images and dimensions of standard signage (ch 4)
https://www.gov.uk/government/publications/traffic-signs-manual
And a couple of suppliers that have dimensions of their wares
https://www.broxap.com/leeds-cast-iron-bollard.html?
http://www.streetfurnishings.co.uk/product-range/traffic-bollards-highway-maintenance-products/
Mr Whippy
Active Member
Nope no thread was started. Just as I get around to this kinda stuff again I end up getting busy with work projects again.
Cheers for the data links!
I still want to make some decent uk roads, this kinda resource would make life easier.
That new Highlands AC track is inspirational, but we need a Dales track haha.
A bit narrower and even faster for those with brass balls!
Cheers for the data links!
I still want to make some decent uk roads, this kinda resource would make life easier.
That new Highlands AC track is inspirational, but we need a Dales track haha.
A bit narrower and even faster for those with brass balls!
Pixelchaser
Well-Known Member
my track has its terrain taken from the Yorkshire dales. and is based on the fundamental design rules of such roads. A & B class for now
Pixelchaser
Well-Known Member
well I'm just laying fictional roads and creating more of a free roam arena suitable for racing too, eventually. pics all through this thread.
http://assettocorsamods.net/threads/rtb-race-track-builder.218/page-6
http://assettocorsamods.net/threads/rtb-race-track-builder.218/page-6
LilSKi
Well-Known Member
This is the site. You don't download from there but you can get the info on the data and it will send you to the place the holds it. Sometimes though the project isn't complete yet and also sometimes it just give you and email or phone number of who holds the data so it may not be free.
https://coast.noaa.gov/inventory/
This is the site where most of the data seems to be held. Decent interface to download the tiles you need.
https://viewer.nationalmap.gov/basic/
https://coast.noaa.gov/inventory/
This is the site where most of the data seems to be held. Decent interface to download the tiles you need.
https://viewer.nationalmap.gov/basic/
Pixelchaser
Well-Known Member
the coverage stops 500 metres before the track I want to create. Doh !Share this page
Support the site
