Creating terrain heightmaps

  • Just an update on my experience with high res terrain mesh in the USA.

    I've managed to get rid of the discontinuities / "cliffs" between adjacent 1 x 1 degree download areas by geoconverting a 2 x 2 block of them together, using a TMC file whose NW co-ordinates are the NW corner of the northwestern-most 1x1 degree area and whose SE co-ordinates are the SE corner of the southeastern-most 1x1 degree area. I put all the IMG files and their corresponding TFW files together in the imput_aerial_images folder. This eliminates all the masked tiles on the boundaries and pushes the problem to the outer boundary which can, of course, be as large as you like. The block could be 2 x 2, 4 x 4 or whatever.

    I've also attached a couple of comparison screenshots below. They were taken on the Elwha River SW of Port Angeles, Washington State. The top one is with the high res terrain mesh and the bottom one is with the default mesh. Note in particular the ravine, which is completely absent in the default mesh screenshot and very well defined in the high res shot.

    Hey Ian, this looks fantastic! Nice job, and glad you were able to figure out how to do it. It makes a huge difference in certain areas. One of the nice side-effects of this is that it also makes the satellite orthos look better, because they get properly stretched over the terrain shapes instead of being distorted over flat ground.

  • Hey Ian, this looks fantastic! Nice job, and glad you were able to figure out how to do it. It makes a huge difference in certain areas. One of the nice side-effects of this is that it also makes the satellite orthos look better, because they get properly stretched over the terrain shapes instead of being distorted over flat ground.

    Thanks qwerty - and many thanks to you also for figuring this out in the first place. That's true about the orthos. There are so many ways in which it makes a difference, particularly if you fly low ..... no more sloping lakes, rivers and sea, no more angles on mountain ridges, fine knobbles, ravines and hummocks modelled etc etc.

  • please tell us how to do it

    Hi again Chris

    qwerty42 's video on page one of this thread is definitely your first port of call. It is very thorough and carefully explained.

    Regarding The Hood River 1x1 degree area (n46w122) .....

    Use the TMC file that I gave you for the geoconversion. Put the TFW file that I gave you into the "input_aerial_images" folder along with the TIF file that you get when you save the IMG file using QGIS. (Add layer > add raster layer > save as etc.) The TFW and TIF files should have the same name.

    Cheers

    Ian

  • excuse my ignorance but i havent got a clue on how to find threads and stuff if tried finding it but no luck

    you said you found an easier way for australian mesh can you explain pleas

    I-74790 cpu win10 gtx1080TI 32 gb ram 2x1TB ssd:P

  • Just an update on my experience with high res terrain mesh in the USA.

    I've managed to get rid of the discontinuities / "cliffs" between adjacent 1 x 1 degree download areas by geoconverting a 2 x 2 block of them together, using a TMC file whose NW co-ordinates are the NW corner of the northwestern-most 1x1 degree area and whose SE co-ordinates are the SE corner of the southeastern-most 1x1 degree area. I put all the IMG files and their corresponding TFW files together in the imput_aerial_images folder. This eliminates all the masked tiles on the boundaries and pushes the problem to the outer boundary which can, of course, be as large as you like. The block could be 2 x 2, 4 x 4 or whatever.

    The technique I use for high res terrain mesh is to always compile a wider area in a low enough level for smooth transition with default terrain.

    Below is an example illustrating this with a local 5m dem mesh in Courchevel (France): I compiled in this case a Level 9 tile and took care to reduce the extent of the Level 12 compilation by at least 1 tile.

    The table below shows my understanding of AFS2 tile sizes and corresponding resolutions :

    Based on this, a 5m dem mesh may be compiled up to Level 12 in AFS2.

    The outer zone compiled in Level 9 gives a resolution of approx 38m/pixels, which usually allows for a smooth transition with the default mesh, provided that it's not too inaccurate.

    I haven't found a parameter yet to minimize mesh compilation rounding, so the result of 5m dem compilation is not as good in AFS2 as in FSX/P3D, but still acceptable in most cases.

    For better results, a specific mesh can be compiled as a TGI, like below (LFLJ runway integration)


    Cheers

    Antoine

    Config : i7 6900K - 20MB currently set at 3.20GHz, Cooling Noctua NH-U14S, Motherboard ASUS Rampage V Extreme U3.1, RAM HyperX Savage Black Edition 16GB DDR4 3000 MHz, Graphic Card Gigabyte GeForce GTX 1080 8GB, Power supply Corsair RM Series 850W, Windows 10 64 bit.

  • Thanks Antoine - that's interesting

    So far I haven't looked at what happens at the transition between high res and default. I've just stayed within my high res area.

    A couple of questions:

    (1) Where are you getting your high res terrain data for France?

    (2) Where did you get those very fine looking buildings?

    France has a lot going for it in terms of scenery creation .....

    (1) It has a wide variety of dramatic scenery.

    (2) The OSM data is very good - particularly for trees - but even buildings aren't too bad. In the USA I have to create my own tree cultivation using the ScenProc "texture filter editor".

    (3) The aerial images are good

    ..... so if I could get high res terrain data I might spend some time in France too!

    Ian

  • i dont know if this is the right"thread" but has someone got .spc for cultivation that puts in all trees ,airport buildings,all residential coverage ,bridges,roads highways and skyscrapers? please

    i cant code and would like to modify my current template_ver6 spc template

    thank you

    chris gard

    coxo2436

    I-74790 cpu win10 gtx1080TI 32 gb ram 2x1TB ssd:P

    Edited once, last by coxo2436 (January 20, 2019 at 4:59 PM).

  • i dont know if this is the right"thread" but has someone got .spc for cultivation that puts in all trees ,airport buildings,all residential coverage ,bridges,roads highways and skyscrapers? please

    i cant code and would like to modify my current template_ver6 spc template

    thank you

    chris gard

    coxo2436

    Chris

    I got this SPC (attached) Template_ver6.txt originally from crispy136. It does most things - but probably not bridges. Bear in mind that coverage depends on the quality of the OSM data. In the USA and Australia you'll probably need to use the ScenProc "texture filter editor" to get good tree coverage. I explain how to do that here: Cultivation from photoscenery?. In the light of experience since I wrote that post I would now say: (1) You can do areas as big as level 10 quite successfully. (2) To get the BMP image I set the FSET download resolution to 5. (I don't think (?) you can use Aeroscenery to get the image because Aeroscenery doesn't give you an INF file - but maybe there is some way of doing it? Others will no doubt know.); (3) Setting the "buffer size" and "minimal feature size" both to "2" gives more precise results; (4) I haven't checked that the 2 SPC config scripts work with the latest version of ScenProc. I've been too lazy to update them and use an earlier version of ScenProc just for that purpose.

    Ian

    Edited 9 times, last by Ian C (January 21, 2019 at 1:46 AM).

  • So far I haven't looked at what happens at the transition between high res and default. I've just stayed within my high res area.

    A couple of questions:

    (1) Where are you getting your high res terrain data for France?

    (2) Where did you get those very fine looking buildings?

    Hi Ian,

    The potential issues with mesh compiling are usually border effects. The higher the resolution, the worst the cliff effect at interface between meshes from different sources, since the heightmap has to correct any discrepency within a shorter distance.

    My 5m dem sources are from a commercial supplier, thus not available for free.

    As well, the Courchevel scenery is a quick and dirty porting test from a commercial product for FSX, nothing that could get distributed as such I'm afraid.

    Cheers

    Antoine

    Config : i7 6900K - 20MB currently set at 3.20GHz, Cooling Noctua NH-U14S, Motherboard ASUS Rampage V Extreme U3.1, RAM HyperX Savage Black Edition 16GB DDR4 3000 MHz, Graphic Card Gigabyte GeForce GTX 1080 8GB, Power supply Corsair RM Series 850W, Windows 10 64 bit.

  • OK - thanks for your reply Antoine.

    Before I sorted out the masking issues along the boundaries between high res terrain areas I had "cliffs" that looked to be nearly 100 m high in mountainous districts - so yes, the discrepancy between default and high res terrain can be very considerable. (There was a strip of default terrain between the high res areas.) There is some even higher res data available in the USA (1/9 arcsec) but it's only available in small patches. However I was thinking that, if I do decide to use it, the discrepancy between 1/9 and 1/3 arcsec terrain shouldn't be too great - probably no more than a metre or so.

    Cheers

    Ian

  • Rather than pure resolution, I'd say the main issue is mesh quality and consistency of coverage.

    DEM provider use algoritms to eliminate outlier measurements. A too strong averaging algorithm tends to round up terrain specific features, while a too weak process leaves erroneous spikes in the DEM.

    As a result, the higher level you compile, the more unwanted spikes you may get in your mesh.

    Even commercial products like Intermap's Next have such issues, like an ugly spike instead of the Matterhorn. Intermap support replied they were quite aware, but since so few people complained so far, they were not going to handle the bug (sic).

    It may read a paradox, but a high resolution mesh is usually much better for rather flat landscapes than for mountainous areas.

    The very high res DEM may outline very specific features in an otherwise rather flat terrain, like for instance depressed trenches of roads/railways, hills, etc. that greatly enhances the countryside of a place like Belgium or northern France, while a standard 19m mesh leaves everything flat.

    High resolution DEM tends to make mountains exaggeratedly spiky, and in some cases causes erroneous artefacts that disappear when reducing the compilation level, a compromise has thus to be found.

    Masks are also a touchy topic in AFS2 and I always try to avoid them when I can do without. I already happened to notice strong border effects when compiling a too small area involving masks.

    If your transition high res to low res is compiled out of the very same source (like the Level 12 to Level 9 interface in my example above), then you shouldn't get a cliff.

    Cheers

    Antoine

    Config : i7 6900K - 20MB currently set at 3.20GHz, Cooling Noctua NH-U14S, Motherboard ASUS Rampage V Extreme U3.1, RAM HyperX Savage Black Edition 16GB DDR4 3000 MHz, Graphic Card Gigabyte GeForce GTX 1080 8GB, Power supply Corsair RM Series 850W, Windows 10 64 bit.

  • Good point - yes - I hadn't thought about noise and spikes in the data. So far I haven't noticed any in the USA but maybe it's not an issue with approx 10 m resolution (1/3 arcsec) and level 12 compilation. (Actually level 11 would probably suffice for 10 m res in theory.) The resolution is high enough to create pleasing mountain features but not so high as to create the problems you mention.

    Cheers

    Ian

  • A short question about terrain modeling: Is it possible to convert a free terrain model of Norway - over to AFS2?

    (DEM/TIFF/OGC WMTS/WMS/png/jpg/svg/GML/SOSI)?

    Norwegian: "Terrengmodellen brukes også i flysimulatorer og til å finne dekningsområder for radar, TV og mobiltelefon."

    Translated: "The terrain model is also used in flight simulators and to find coverage areas for radar, television and mobile phone."

    https://kartkatalog.geonorge.no/metadata/kartv…40-7ec04c0e3918

  • A short question about terrain modeling: Is it possible to convert a free terrain model of Norway - over to AFS2?

    (DEM/TIFF/OGC WMTS/WMS/png/jpg/svg/GML/SOSI)?

    Technically, the TIFF format should be easy to handle... Just open in QGIS, set your TFW and TMC files and there you go.

    Cheers

    Antoine

    Config : i7 6900K - 20MB currently set at 3.20GHz, Cooling Noctua NH-U14S, Motherboard ASUS Rampage V Extreme U3.1, RAM HyperX Savage Black Edition 16GB DDR4 3000 MHz, Graphic Card Gigabyte GeForce GTX 1080 8GB, Power supply Corsair RM Series 850W, Windows 10 64 bit.