[Lindsayts]

This is the first of a series of posts on how to get an accurate world model in a train sim. Now I am not an expert in anyway on cartography and before this new little about the subject, but the following is the summary of much research as I will not let lack of knowledge be the cause of a project to fail.

The rest of the posts will be in the Physics, railways forum.

The projection to do the terrain generation is Transverse Mercator,TM, this was choosen as all high detail maps use Universal Transverse Mercator, UTM, this being a particular case of the more general TM. The mathematics being exactly the same.

TM itself produces a strip of terrain of any length North South, but in the East West direction is limited by scale distortion. The scale in TM being constant in a North South direction but varies in an East West direction. Some idea though on the accuracy expected is the if a scale factor of 0.9994 is used during terrain generation. The East West scale will not deviate by more than 0.04% at distance on aprox 250 kilometres either side of the reference longitude, so a route considerable wider than that could be done in real life. My own work consistently shows the distances as modeled by this model are VERY accurate, in fact I found its almost impossible for long distances on the real earth to be measured with an accuracy equal to the model.

The co-ordinates for TM are in metres so tiles of say 2000 metres square are not a problem. The terrain is still tiled so little change in OR apart from the co-ordinate conversion will need to be done. A freely distributable application is readily availible for individual conversions. This app uses a WELL documented library to do the actual conversions, This should simplify implementation. The library and app like I said are freely distrbutable on the condition the authors are cleared of any responsibilty for any thing that may happen when using the library/app.

The mathematical methods for both the spherical and the ellipsoidal versions of Transverse Mercator have been published and are readily avialible if the OR devs wish to go down that path. The more accurate form is the ellipsoidal, the mathematics are quite a bit more complex but as its only used on a large scale during terrain generation which is not a real time process, this should not be a problem.

Past reactions on this subject are along the lines of "god thats complicated" hopefull readers will take the time to ask if anything is not clear.

Any questions please ask.



Lindsay

[Genma Saotome]

Two questions:

Why not go with UTM?

Why should the tiles be square? Couldn't the boundaries on the north and south of the tile be set at different lengths, correct for their latitude? IOW, why does the game world need to be 2d? Of course difficulty to implement for little gain is a valid answer... but if that is not the case then why not use 3d? It's just math after all.

[captain_bazza]

I wonder, are we really talking aboot two issues here:

1] what is required for a user-friendly route editor;

2] which DEM data format(s) are desirable to achieve realistically sculptured terrain,

and can it be packaged for ease of distribution, and installation by the end user?

((At some stage I'm going to start a discussion topic about Outerra Anteworld, having purchased, for US$15.00, the alpha version. There are some interesting ''pro's & con's'' involved.))

Cheers Bazza.

[Lindsayts]

Genma Saotome, on 30 August 2014 - 03:27 PM, said:

Two questions:



Why not go with UTM?

UTM (Universal Transverse Mercactor) divides the world into distinct 6 degree zones, because each zone has its own central meridian and the projection is of a compound surface onto a flat plan. The boundries between Zones DO NOT match. So one cannot build a route across a zone boundry. Transverse Mercator is the same projection the only difference is the author picks the central meridian to suite the route,as the zone is around 700 kilometres wide its unlikely a single group of route developers will ever reach the boundry.

Quote

Why should the tiles be square? Couldn't the boundaries on the north and south of the tile be set at different lengths, correct for their latitude? IOW, why does the game world need to be 2d? Of course difficulty to implement for little gain is a valid answer... but if that is not the case then why not use 3d? It's just math after all.

The co-ordinate grid for UTM and TM is to all intents and purposes 1000 metre squares. The projection automaticly compensates for increases in latitude. If you look at UTM maps such as the 1 in 100,000 series one will find the UTM co-ordinate being 90 degree squares does not match the meridians as these of course converge towards the poles.

Lindsay

[captain_bazza]

Yeah, it's a real pucker up job, this far south, mate.

Cheers Bazza.

[Lindsayts]

Genma Saotome, on 30 August 2014 - 03:27 PM, said:

Why should the tiles be square? Couldn't the boundaries on the north and south of the tile be set at different lengths, correct for their latitude? IOW, why does the game world need to be 2d? Of course difficulty to implement for little gain is a valid answer... but if that is not the case then why not use 3d? It's just math after all.

Answer no 2,

You sound like you are talking about another method to do terrain thats used by some flight sims. 3DEM is used directly, the tiles being scaled to the correct dimensions for there latitude and Longitude. I have experimented with this method and in fact its the one I am using in my own sim.

Lindsay

[Lindsayts]

captain_bazza, on 30 August 2014 - 05:31 PM, said:

I wonder, are we really talking about two issues here:

These two questions are really outside scope of the world model but.......

Quote

1] what is required for a user-friendly route editor;

This of course is the killer, considering though the MSTS editor and tools, the first consideration MUST be for OR's editor to be RELIABLE.

Quote

2] which DEM data format(s) are desirable to achieve realistically sculptured terrain,

and can it be packaged for ease of distribution, and installation by the end user?

((At some stage I'm going to start a discussion topic about Outerra Anteworld, having purchased, for US$15.00, the alpha version. There are some interesting ''pro's & con's'' involved.))

Cheers Bazza.

DEM data is just that data, the format is not significant as long as the Data is good.

A number of points need to be considered, first is for most of the world 3DEM is the best one that is availible, so this has to be the lowest common denominator.

The second point is the the DEM data has been created by repeated scans of radar and is actual NOT the height of the terrain UNLESS the terrain is bare of vegitation. For most places though the numbers in the DEM data reflect the height of the vegitation. In forested areas such as the very tall eucalypt forests in the great dividing range in Victoria, Australia, its not uncommon to find the 3DEM data to 50 to 80 metres higher than the real terrain due to the many giant trees.

Lindsay

[That Genset Foamer]

I've noticed that with urban 30m SRTM data--it'll "warp" at points to the skyline and not the ground, or in the case of a couple blimp hangars in my general vicinity it'll produce two similarly-proportioned mountains. For comparison 10m NED data goes down to soil and "removes" the houses altogether, and doesn't produce the aforementioned hills.

[Genma Saotome]

I'm just asking questions, not meaning to cause any trouble.

WER projecting DEM into a 3d grid... seems to me the DEM data can be made to fit a 3d "tile" that isn't flat or perfectly rectangular... that's why I asked. Laying out road and track already requires vertical adjustments. I guess I should have phrased it differently... something like "Why wouldn't we do a real 3d world?". Ignorance comes up w/ such questions.

Lindsay, since you are going down that path yourself, why recommend something else for OR?
:sign_thanks:

[Lindsayts]

Genma Saotome, on 30 August 2014 - 10:44 PM, said:

I'm just asking questions, not meaning to cause any trouble.

WER projecting DEM into a 3d grid... seems to me the DEM data can be made to fit a 3d "tile" that isn't flat or perfectly rectangular... that's why I asked. Laying out road and track already requires vertical adjustments. I guess I should have phrased it differently... something like "Why wouldn't we do a real 3d world?". Ignorance comes up w/ such questions.

This is a relatively complex subject, questions are ALWAYS welcome.

DEM data is supplied in geocentric co-ordinates, these are of course basicly trapezoid in shape, during the terrain generation process a great deal of complex co-ordinate conversion is going on, converting the data to a flat rectangular world.

The problem with doing a 3d world is the amount of processing power needed to display it. This is THE great beauty of a tiled flat world, the small amount of both CPU and GPU power required to display it, leaving more grunt availible for rolling stock and scenery.

Quote

Lindsay, since you are going down that path yourself, why recommend something else for OR?
:sign_thanks:

Both methods will be being presented, its up to others to determine which is to be used. The advantage of the tiled Transverse Mercator model being presented first is a good deal less work is likely to be required to intergrate it in OR.

Lindsay