Elvas Tower: Map projectoions for terrain simulation - Elvas Tower

Jump to content

Page 1 of 1
  • You cannot start a new topic
  • You cannot reply to this topic

Map projectoions for terrain simulation What projections are suitable for a train sim

#1 User is offline   Lindsayts 

  • Superintendant
  • Group: Elite Member
  • Posts: 1,787
  • Joined: 25-November 11
  • Gender:Male
  • Country:

Posted 11 January 2018 - 10:07 PM

Our host Dave Nelson has ask me to comment on this, IF ANYTHING IS UNCLEAR PLEASE ASK.

Now I will say right from the start this is a real thorny problem, there is and will be NO perfect solution. The main problem being the earth is (sort of) spherical (Note 1) and its NOT possible to accurately project such a surface onto a flat plain.

Note 1: The earth is NOT a perfect shape, its a slightly flatened sphere, the diameter at the poles being LESS than at the equator, the radius of the earths surface from the earths centre is also irregular, the portion of the earths furthest from the centre, ie the highest point, is actually a mountain in northern South America.

There is 3 projections one can consider using, Transverse Mercator, Universal Transverse Mercator and Lambert Conformal Conic Projection. All three will produce flat maps in which the distance AND direction both can be measured with a SPECIFIED degree of accuracy..

Transverse Mercator ™ and Universal Transverse Mercator (UTM) are mathematical EXACTLY identical, the only difference is the way there origin is specified. The origin in both projections is the meridian in the centre of the area the map covers. In UTM this is actual part of the maps specification, the central meridian is every 6 degrees starting at 3 degrees, so the progresion is 3,9,15,21 etc, the map area is 6 degrees wide ie plus or minus 3 degrees from eh central meridian. The maps so produce have a KNOWN built in error which for all intents and purposes is negligble. The main problem is using this for a simulation is is if the area one is modilling fits into on of the 6 degree bands everything is great if not there is little one can do. TM solves this issue, As I said both TM and UTM use EXACTLY the same mathematics, TM allows one though to specfiy BOTH the central meridian and the projections width, ie the central could be 3, 9, 15, 134, 75.3 etc. In reality one would use as the central meridian, the closest Meridian to the routes East West centre.

The third proection is Lambert Conformal Conic Projection (LCCP), In the Mercator family of projections the surface one is projecting on is a parrallel sided cylinder that touchs the earths surface ONLY at the projections central meridian. in LCCP the projection surface is a cone whose apex co-insides with the earhs axis, the cone cuts the earths surface at 2 points usually one uses integer latitudes. The area mapped comprises everything between the two parralells plus some distance above the top parrallel and some distance below the bottom. I have personally done little work on this as I chose TM as the projection to be used, but as I said this will also produce an a map which is accurate to tightly specified limits.

One thing worth considering is to actually get the projection accurate enough for a ACCURATE train sim, the earths true shape needs to be taken into account, the only projection that has the full mathematics published to allow this (as ar as I know) is the Mercator family of projections.

I will bring up the point there is another way to generate true accurate terrain where there is NO limit to the worlds size as there is for both TM and LCCP, terrain is still tiled. The method uses rectangular array data as most geo mapping data is supplied and simply scales the tile the correct amount just prior to it being displayed, there is at least 2 flight sims that use this method. There appears to be little speed disadvantage as such tile scaling i built into most current GPU's.

Lindsay

#2 User is offline   Lindsayts 

  • Superintendant
  • Group: Elite Member
  • Posts: 1,787
  • Joined: 25-November 11
  • Gender:Male
  • Country:

Posted 11 January 2018 - 10:25 PM

Now if anyone wants to play around with the various map proections, its quite difficult to beat http://earth-info.ng...GandG/geotrans/,

its an application the converts positions between a good number of mp projections. Documentation is excellent.

Lindsay

#3 User is offline   Lindsayts 

  • Superintendant
  • Group: Elite Member
  • Posts: 1,787
  • Joined: 25-November 11
  • Gender:Male
  • Country:

Posted 11 January 2018 - 10:47 PM

Note: I am familiar with QGIS, the most serious problem I have found with this (and other such software) is you first need to fully understand the data and then what one wishs to do with it, this is no where near as simple as it seems.

Lindsay

#4 User is offline   dajones 

  • Open Rails Developer
  • Group: Active Member
  • Posts: 266
  • Joined: 27-February 08
  • Gender:Male
  • Location:Durango, CO
  • Country:

Posted 12 January 2018 - 06:15 AM

The full math for many projections can be found in https://pubs.usgs.go...395/report.pdf. But it makes more sense to use the proj.4 API (http://proj4.org) instead of implementing any of it.

Doug

#5 User is offline   Garry 

  • Hostler
  • Group: Active Member
  • Posts: 79
  • Joined: 09-March 15
  • Gender:Male
  • Simulator:Open Rails + TSRE5
  • Country:

Posted 12 January 2018 - 07:12 AM

Have you looked at this thread under route editing?
http://www.elvastowe...map-projection/

Garry

#6 User is offline   Lindsayts 

  • Superintendant
  • Group: Elite Member
  • Posts: 1,787
  • Joined: 25-November 11
  • Gender:Male
  • Country:

Posted 12 January 2018 - 09:39 AM

The idea behind this thread is that Goku has included some kind of more accurate projection in his route editor, its this writers hope we can get some intelegent discusion going on projections and terrain creation to allow us, first to use Gokus route editor to best advantage, second to allow Openrails main developers a better insite into what projections can be used and what opensource tools are around to help with terrain creation.

As one has already mentioned there is a number of open source projects around claiming good accuracy so the mathematics for the conversion process is a none issue. How Goku is implementing this will be more important than the actual projection used.

A point I will make is that Goku's is producing some tools to for use in OR as a private venture, there is NOTHING wrong with this and I hope his work is not taken as any kind of threat. Some people myself included actually like to do such work alone.

Lindsay

#7 User is offline   Lindsayts 

  • Superintendant
  • Group: Elite Member
  • Posts: 1,787
  • Joined: 25-November 11
  • Gender:Male
  • Country:

Posted 12 January 2018 - 12:43 PM

As previously mentioned anyone playing around with geodetic coordinate conversions is strongly recommended to try Geotrans. This is availible for both Windows and Linux and the application clearly displays what data is required for a conversion. This makes research in how to use any projection much easier I found.

#8 User is offline   darwins 

  • Apprentice
  • Group: Switchman
  • Posts: 10
  • Joined: 25-September 17
  • Gender:Male
  • Simulator:Open Rails
  • Country:

Posted 15 January 2018 - 09:49 PM

I am not sure that I will be able to play around with this, but I welcome any more accurate map projection.

Unlike a flight simulator, few of us travel far enough round the globe in rail simulators to be aware of the curvature. (Was anyone really starting the trans-siberian railway??)

What we would like is something that will give us reasonably accurate distances (E-W and N-S) and heights, as well as not being horribly skewed. I did recently attempt to start some routes using the de-skew fix the later version of Demex. Whilst things did look better than a standard MSTS projection within the short distance attempted (about 30 km) I was finding that some markers had been squeezed together and others stretched apart. I tried some texturing from maps and found that I could not keep to a constant horizontal and vertical scale when converting them to terrain textures.

So it is very exciting to hear that Goku has found "any" better way to project the surface of the earth.

Please, let there also be an easy way to project maps onto it as textures... (Trying to build older often closed routes for me means trying to use some older maps with strange imperial units and often no obvious grid. So cutting them into 200m squares can be slow.)

I am looking forward to follow developments in this area.

Page 1 of 1
  • You cannot start a new topic
  • You cannot reply to this topic

1 User(s) are reading this topic
0 members, 1 guests, 0 anonymous users