Hello everyone,

one question that is on my mind for a couple of years now, is the amount of data needed to create race tracks. The current gold standard for racing simulations is the terrestrial laserscanning, which provides a quick technique to aquire large amounts of geometrical (and radiometrical) data at a very high accuracy. This comes at the cost of big data sets and a complex work flow to filter and clear the data to the customers needs. At this point the real work for the developers of digital models as used in racing games really only starts. A slightly less complex solution would be the use of UAV photogrammetry supplemented by terrestrial photogrammetry, which also creates large point clouds but is faster while slightly less accurate. Still both approaches need trained professionals and come with several challanges concerning among others the stable reference frames needed to create a correct representation of the real world structures. Both photogrammetry and laserscanning can be used complementary.
Due to the effort involved with both techniques, they might become expensive. However geodata sources are in fact more and more available, if at a very reduced information density. Some sources are even open to the public without any additional costs involved. With the view from my personal bubble I think that the following data sources might also be a good starting point for the development of race tracks with a basis on real life data.

Airborne Laserscanning provides elevation data over large areas. Unlike laserscanned racetracks with millions of points, the public data only provides 4 to 10 points per square meter with an irregular distribution and an accuracy of 30cm in position and 15cm in height. The point clouds are however processed to give rasterized data with 1m spacings across large landscapes as digital terrain models or digital elevation models. These are low-pass filtered representations of the real terrains. While it is too little in terms of accuracy and points to replicate the real-life appearance, it gives a good generalized representation of the topography in relative high detail. These data sets are provided in .csv ASCII files (terrain models) or in LAS-files and LAZ-archives (point clouds).

Especially when combined with digital orthophotos (especially true-orthophotos) from flight campaigns, they give a good impression of how the world looks. Orthophotos can be obtained in .jp2 format or as an online service via OGC-WMS standard.
For buildings especially 3D building models in the LOD2. LOD2 are untextured, where is LOD3 are textured 3D models in the CityGML format, sometimes also provided as an OGC-WFS service.




The German state of North Rhine-Westphalia provide several datasets in open data under the so-called Data licence Germany - Zero - Version 2.0, which allows commercial of these data sets, so their use as a basis for an own commercial product is allowed (and encouraged). I put the links below, however they are in German.
terrain model: https://www.bezreg-koeln.nrw.de/brk_internet/geobasis/hoehenmodelle/digitale_gelaendemodelle/gelaendemodell/index.html
building models: https://www.bezreg-koeln.nrw.de/brk_internet/geobasis/3d_gebaeudemodelle/index.html
digital orthophotos: https://www.bezreg-koeln.nrw.de/brk_internet/geobasis/luftbildinformationen/aktuell/digitale_orthophotos/index.html
Data licence Germany - Zero - Version 2.0: https://www.govdata.de/dl-de/zero-2-0

I expect one problem off all such geodata is the scale as the ETRS89/UTM coordinates (and other geodetic/geographic coordinate systems) used for georeferencing are at a variable scale. It would probably best to transform an interesting data set into a localized cartesian coordinates first.

Using such data sources does not mean, you can get a 3D-race track from just combining a few data sources, but I believe they can be a good ground to start working with, by identifying interesting areas, getting a raw geometric representation of the real world and thus improving and densifying the data by designing the missing elements and pieces into an interesting unique yet also realistic track. Especially textures would need to be created, since the available public/open data files at best have 10cm per pixel resolutions (for privacy reasons).

Best regards

EDIT: I am happy to answer your questions here. I know it's a big wall of text and partly without proper formulation.

I could imagine that a combination of techniques could speed up the track-modelling process. E.g. A traditional laser-scan of the track surface and kerbs dropped into models using this airborne technique. You don't need centimetre accuracy for buildings and surrounding terrain, so applying the most appropriate technique for each aspect of the model could save time and resources.

This reminds me of when Google announced their 3D Maps Platform where you can use their 3D map data. It would be a good starting point to design a suitable track layout, but wouldn't be enough for an accurate end-product.

The resolution of airborne techniques is not currently adequate to properly model the track surface, the part that actually matters in a simulation. Minute variations in the tarmac couldn't be modeled accurately enough with the technique you presented here. You'd still need someone to laser-scan the track surface.

It's so cool that this data is available to use though, I just wish there was more time in a day to make use of it myself.

Quote from mbutcher :

I could imagine that a combination of techniques could speed up the track-modelling process. E.g. A traditional laser-scan of the track surface and kerbs dropped into models using this airborne technique. You don't need centimetre accuracy for buildings and surrounding terrain, so applying the most appropriate technique for each aspect of the model could save time and resources.

I completely agree! Track surfaces would need to have either a detailed modelling from scratch or, if available, a surface scan like it is possible with a profile scanner on a multi-sensor vehicle for mobile mapping. City agencies for road building and civil engineering are using such vehicles to generate maps and data sets to assess the conditions of the roads.
In theory the UAV photogrammetry could also work, however due to the homogenous, featureless look of the road surfaces I don't expect it to detect the undulations that well, neither with the fidelity nor with the accuracy.

Quote from mbutcher :

The resolution of airborne techniques is not currently adequate to properly model the track surface, the part that actually matters in a simulation. Minute variations in the tarmac couldn't be modeled accurately enough with the technique you presented here. You'd still need someone to laser-scan the track surface.

Yes, no doubt, the data sets I presented are more appropriate to model topography and and the general shape of the surface, as well as the building. The accuracy of the buildings in the datasets is already close to decimeter accuracy, as they are derived from the real estate cadaster (data quality and fidelity of which is a story for another time...) but with simplified geometry.

Great topic.
It would be possible to create a virtual racetrack in parts of the world where there isn't a track now. Choosing some interesting or tricky topography from the real world and then building your track at that location could make for some great racing.

This is an example of what photogrammetry alone is capable of.

That's really quite good! Well done.