FSXA Terrain mesh problem



Does anyone know how to remove the spikes which are seen here clearly on the top of a cliff? I am aware that they are caused by over-sampled mesh, but i want to remove them without having to lower the LOD.
 
First, where is that and is the coast line default or custom? The default coastline never really matches DEM expecially ones with a higher resolution. Then, in some parts of the world even coastlines that are traced in GE are in a mismatch to a DEM of up to 150 meters. This becomes obvious when one side of an island is totally flat (0m) at the edge and the other one is cut off like with a knife. If I observe this I usually move the hole poly of the island in SBX to make it fit the DEM.

Second, if above can be ruled out, the DEM has to be analyzed.
Did you have a look at the source file with 3DEM or something similar?
If I had to solve this by working the DEM over, I'd have a look at the source to see what's really there, and if it needs to be worked over I'd correct the heightfield with the Daylon Leveller, convert it back to tiff and resample it again.

You might be able to correct the spiky edges with sloped flatten polygons in SBX et al, but that is awkward, at least in my opinion.

HTH
Mark
 
Hi Mark,

The coast-line is custom made via SBuilderX, following the real thing, pretty much ruling out your first suggestion. I'll give a look at the DEM data in Global mapper or 3DEM and see if it needs to be worked on....
 
Does anyone know how to remove the spikes which are seen here clearly on the top of a cliff? I am aware that they are caused by over-sampled mesh, but i want to remove them without having to lower the LOD.
Hi Alex:

A few questions might help better determine what factor(s) may have contributed to the anomaly at the edge of your vector data area: :confused:

* What was the horizontal and vertical resolution of the custom terrain mesh DEM source data ?

* What was the DEM source data 16-bit ...or 32-bit ?

* What was the DEM source data *.BIL ...or *.GeoTiff ?

* What was the custom terrain mesh LOD output requested from FSX SDK Resample in your *.INF file ?

* What was the % compression of custom terrain mesh output requested from FSX SDK Resample in your *.INF file ?

* What was the FSX Terrain Mesh Complexity slider set at for the screenshot ?

* What was the FSX Terrain Mesh Resolution slider set at for the screenshot ?


You may also wish to post your *.INF file in your reply to allow a more informed interpretation of this scenario by would-be helpers here. ;)

GaryGB
 
Last edited:
Hi GaryGB,

Here are the answers to the questions you requested above in the correct order.

* 1916 x 934

* 16-bit

* TIF File (.tif)

* LOD 9 to 12

* No compression (0%)

* Mesh Complexity: 75

* Mesh Resolution: 5m

And finally the .INF File used to Resample the DEM data. View attachment Elevation.inf


(Cliff side picture of the problem)
 
Last edited:
"* What was the horizontal and vertical resolution of the custom terrain mesh DEM source data ?"
Hi GaryGB,

Here are the answers to the questions you requested above in the correct order.

* 1916 x 934
Hi Alex:


Regarding the question on horizontal and vertical resolution, your reply was the dimensions in pixels of your source DEM GeoTiff image instead of resolution. ;)

The info on horizontal and vertical resolution would be found in the Metadata documentation which accompanied the DEM data set.

Generally, it should indicate what the distance in between elevation data points is on the ground ex: 90 Meters / 30 Meters / 10 Meters etc. ...which would be the "Horizontal" (Longitude-Latitude or X-Y) axis resolution

Also, it may indicate the precision of the data for the elevations of the DEM ...which would be the "Vertical "(Altitude or Z) axis resolution


PS: Thanks for posting the additional view of the cliffside showing the extent of the anomaly. :)


GaryGB
 
Last edited:
Alex, can you post a screendump from 3DEM or GM of the source Geotiff? Maybe with a screen of the properties?

That's a mighty cliff on the last image :eek:

Cheers,
Mark
 
Hi,

Have you also excluded the default water polygon (not only the shore line)?
If so, have you used a properly calibrated map to make the exclude water poly?
Either that or the mesh data are incorrect;
Or else, the island does indeed have such steep cliffs:);
Or else still...?
 
Mark,

Here are some pictures taken from Global Mapper


(Top down view of the area covered with the new DEM data)


(3D view of the cliffs that create the problem)

robystar,

For the main island of Lampedusa, as seen below, i have not excluded the default water polygons.





GaryGB,

My apologies, i thought you meant the pixel horizontal and vertical (width,height) resolution. Unfortunately the DEM data was not accompanied by any Metadata documentation, i'm sure i can find the resolution somewhere in global mapper but i cant remember where. Regarding the location of the problem sorry for leaving you high and dry, the problem is located on the west side of the Italian island of Lampedusa, samos is one of my other projects.

 
Last edited:
Hi Alex:

Thanks for the update on the actual geographic area under discussion here; I've edited my prior post above. ;)

Interesting project areas, with some 'similar looking' rugged coastlines ! :)


[EDITED]

PS: I have downloaded the "30 Meter" Horizontal resolution ASTER GDEM data for Lampedusa, Italy, and will take a look at this scenario in Global Mapper, and via creation of a multi-LOD mesh BGL in FSX SDK Resample, in an effort to better understand how this anomaly might have occurred.

I hope to get some time free to check this out again before the end of this week. :cool:

In the mean time, I'd like to initially ask, when you created the polygon (seen as a "selected" green line with vertex points defining the coastline) in your 2 SBuilderX screenshots above, what BGL output type was it intended to be ?

Is more than (1) SBuilder BGL output type made from that polygon 'active' in FS, thereby allowing its impact to be seen in your FSX screenshots above ?

If so, what SBuilder BGL output types are those which were made from that polygon ?


BTW: If that polygon was used to make a "Flatten", did you assign the points of the polygon (aka the "vertices") individual elevations / altitudes ? :confused:

[END_EDIT]


GaryGB
 
Last edited:
Hi GaryGB,

During the development of Samos i never encountered this sort of a problem. The Polygon was created a while back ( 3-4 weeks ago),BGL output intended was one and it was of a Terrain Vector type.
 
For the main island of Lampedusa, as seen below, i have not excluded the default water polygons.
In the mean time, I'd like to initially ask, when you created the polygon (seen as a "selected" green line with vertex points defining the coastline) in your 2 SBuilderX screenshots above, what BGL output type was it intended to be ?
The Polygon was created a while back ( 3-4 weeks ago),BGL output intended was one and it was of a Terrain Vector type.
The poly done to conform to the outline of the island is nice, but..., it's not going to do what you want unless you adjust the water to match it. You can make the nicest outline in the whole world, but water is still going to take priority when it comes to displaying inside the game. You should be able to make an exclusion poly (or two, three) for the water and use your revised island outline as a hole. That may help clear up some of the mesh anomalies?

And while it's still unknown what the native source resolution was for the mesh file, I'll go with 90m from the display in Global Mapper. Trying to then turn that into some magical 9.5m super mesh is like putting lipstick on a pig. Doesn't really make the pig look anything more than what it is, a pig that's wearing lipstick... :D
 
H Alex:

You may wish to review this excellent tutorial on using SBuilderX to make a photo-realistic scenery of an island, which includes clearly explained and detailed steps for the exclusion of default water polygons and creation of accurate replacement scenery content:


How to create photoreal scenery for FSX - by "Tiberius K."

http://www.flightsim.com/vbfs/showthread.php?250762-How-to-create-photoreal-scenery-for-FSX


[EDITED]

A preceding thread apparently compelled Tiberius Kowalski (aka "Tiberius K." at flightsim.com) to create the above tutorial, points out common errors which can be made when making islands (ex: by incorrect use of Excludes, as well as failure to remove default CVX BGLs local to the area of an island and replace them with custom water polygons and other required content); that thread is here:

http://www.flightsim.com/vbfs/showthread.php?245629-Project-Nauru-PHOTOREAL-(in-progress)/page3


[END_EDIT]


I hope the above linked tutorial helps provide the instruction which some respondents here at FS Developer often seem unable or unwilling to provide, perhaps due to limited time and energy (...or possibly due to limited patience, if it seems that available learning resources and documentation for scenery building tools 'perceived' as being "widely-known" or "readily-comprehensible" ...have apparently not yet been fully explored by a person asking questions here ?) :rolleyes:

IMHO, in the time it takes to compose a post that is (purposely ?) vague or otherwise "actively" unhelpful, one could just as easily have been polite, and simply posted a link to a tutorial somewhere ...that would actually be helpful to someone ! :eek:


Although I personally would not usually envision one's goal as being the creation of a terrain mesh over-sampled more than 1 (...or at the very most '2') LODs beyond the internal horizontal resolution of a downloaded 30 Meter ASTER GDEM terrain mesh source, we can see what may actually prove feasible and reasonably accurate when I get some time to tinker with the Lampedusa GDEM data, and make some terrain mesh BGLs later this week.

However, depending on what degree of detail you might desire for the terrain on Lampedusa, it is possible you may still wish to over-sample to a larger degree, so that you have a higher quad matrix grid resolution of terrain mesh vertices available to allow making higher resolution sloped flattens to correct for terrain anomalies, and to enable display of greater FSX vector detail along coastlines. :idea:

AFAIK, this is what was done in the Playsims Publishing VFR Photographic Scenery Generation X - Volume 8 Scotland South for FSX; the terrain mesh was declared to be "5 Meter" resolution, but IMHO was clearly no better than 90 Meter (or at best, 76 Meter) source data resolution as far as actual terrain "shape" detail seen in FSX at run time, or viewed in FSX SDK TMFViewer.

However, the higher grid resolution of terrain mesh vertices enabled by the 5 meter mesh in that add-on scenery package were may have been made available for whoever wished to implement higher resolution vector flattens for terrain correction.


In the mean time, to ensure a more predictable preview of what you and your intended end users may see when your add-on scenery is loaded and displayed in FSX, I'd recommend using the following settings for all inspection flights and screenshots within FSX:


* FSX Terrain Mesh Complexity slider set at 100%

* FSX Terrain Mesh Resolution slider set at no more than 1 increment farther to the right (...a smaller Meters interval between elevation data points) than the maximum LOD resolution of any output terrain mesh BGL you are using


Hope this helps a bit more ! :)

GaryGB
 
Last edited:
Hi Alex:

I only had a moment free today to make a FSX Terrain mesh from the downloaded ASTER GDEM2 "30 Meter" resolution 16-bit GeoTiff DEM data. ;)


With LOD=Auto in the *.INF file, FSX SDK Resample opted to initially make LODs 1 to 10.

A MAX LOD of 10 correlates with 38.2 M Meters between elevation data points in the output BGL, and would likely be too low of a resolution to provide sufficient terrain vertices for FSX to display a custom high detail coastline with polyline / polygon inter-vertex distances set at what, IIUC, is 20 Meters by default in SBuilderX.

My next test will be to output with some degree of oversampling to increase the terrain grid resolution prior to testing its display in FSX at run time.


Hope this info might still be of interest to you with your Lampedusa project. :)

GaryGB
 
Last edited:
the terrain mesh was declared to be "5 Meter" resolution, but IMHO was clearly no better than 90 Meter (or at best, 76 Meter) source data resolution as far as actual terrain "shape" detail seen in FSX at run time, or viewed in FSX SDK TMFViewer.
I don't know where you get your facts.

How about:

 
Hi Gary,

I believe i have found the cause of the problem, its the "30 Meter" ASTER GDEM which unfortunately isn't accurate enough. When i add the coastline polygons over the DEM it actually cuts through the elevation data causing the spikes, if i extend the polygons out in to sea the problem disappears.



The problem with extending the polygons outwards (as a fix) is that it's not actually following accurately the profile of the island, and so causing other dilemmas.

Another observation i made is, if you over-sample the DEM data the spikes increase, on the other hand, if you lower the LOD the spikes are less in number, bigger in size and the cliffs actually back away 2-4m from the polygon border line.

 
When i add the coastline polygons over the DEM it actually cuts through the elevation data causing the spikes, if i extend the polygons out in to sea the problem disappears.
When you made your polygon in SBuilderX which sat service were you using? If Google Earth, there may be a slight offset. As to why, I dunno, but I did some comparing a good while back for things like airports and using a GE background to do airport work and then putting the airport on top of photo scenery would show a slight difference. If I had used a MS Virtual Earth/Bing?whatever they call it nowadays background, then the airport would align with the photo scenery.

So I'm just saying, if you used GE don't take it as 1000% accurate.
 
AFAIK, this is what was done in the Playsims Publishing VFR Photographic Scenery Generation X - Volume 8 Scotland South for FSX; the terrain mesh was declared to be "5 Meter" resolution, but IMHO was clearly no better than 90 Meter (or at best, 76 Meter) source data resolution as far as actual terrain "shape" detail seen in FSX at run time, or viewed in FSX SDK TMFViewer.

However, the higher grid resolution of terrain mesh vertices enabled by the 5 meter mesh in that add-on scenery package were made available for whoever wished to implement higher resolution vector flattens for terrain correction.
I don't know where you get your facts.
As stated above, Playsims Publishing VFR Photographic Scenery Generation X - Volume 8 Scotland South for FSX ...actual terrain "shape" detail seen in FSX at run time, or viewed in FSX SDK TMFViewer, IMHO was clearly no better than 90 Meter (or at best, 76 Meter) source data resolution as far as actual terrain "shape" detail; that personal observation and personal opinion is not based solely upon LODs listed in TMFViewer, but from direct observation of actual terrain "shape" detail while flying FSX within specific areas.

Your TMFViewer screenshot above does show multiple loaded terrain mesh BGLs with file names from which one might infer they are from Playsims Publishing VFR Photographic Scenery Generation X - Volume 8 Scotland South for FSX.

And of course, the numeric LOD listings in your TMFViewer screenshot above with the loaded BGLs do indeed list LODs ranging from 4 to 13... on what I presume is your own installation, which from which you may be able to identify file versions and file dates.



FYI: To date, I have not flown and inspected more than a specific (small) geographic area from that package, which locally displayed a terrain "shape" detail flying in FSX which I would interpret as being clearly no better than 90 Meter (or at best, 76 Meter) source data resolution.


When that same area was subsequently flown with a 3rd party terrain mesh BGL declared to be 4.75 Meters resolution, FSX displayed a terrain "shape" detail which I would interpret as being generated from least 4.75 Meter source data resolution ...and output to BGL at the declared 4.75 Meter resolution.



I am away from my main FS computer for the weekend, but would be glad to further discuss this issue with you (in a separate thread elsewhere at FSDeveloper) upon my return, and when I again have access to the computer with an installation of the add-on scenery in question.

Perhaps then we can compare results at designated areas with ex: Ordnance Survey GCPs, on-site photos etc., and also compare file versions / file dates from our respective installations of the Playsims Publishing VFR Photographic Scenery Generation X - Volume 8 Scotland South for FSX. :)


BTW: I believe Alex has now endured (2) inappropriate incursions of off-topic digressions in this thread; so IMHO, the right thing to do at this point, is allow this thread to continue on the course the OP might have wanted. ;)


PS: < Oh what the heck... sorry Alex: lets make it (3) inappropriate incursions for some comedy relief ! > :D

[Digression_Mode_On]

George:

What I didn't state above, was that I also observed multiple anomalies and IMHO, serious flaws in the terrain "shape" detail flying in FSX in specific areas with the Playsims Publishing VFR Photographic Scenery Generation X - Volume 8 Scotland South for FSX "5 Meter" terrain mesh loaded. :eek:


This was confirmed by 2 other observers with separately purchased installations of that scenery package, 1 of whom is actually quite familiar with the area in question from low altitude flights and on ground observation in real life.


"Lucky we didn't say anything about the dirty knife" :p

http://www.youtube.com/watch?v=61k1xpiKwEU

http://en.wikipedia.org/wiki/The_Dirty_Fork

[Digression_Mode_Off]


GaryGB
 
Last edited:
When you made your polygon in SBuilderX which sat service were you using? If Google Earth, there may be a slight offset. As to why, I dunno, but I did some comparing a good while back for things like airports and using a GE background to do airport work and then putting the airport on top of photo scenery would show a slight difference. If I had used a MS Virtual Earth/Bing?whatever they call it nowadays background, then the airport would align with the photo scenery.

So I'm just saying, if you used GE don't take it as 1000% accurate.
In my case i used Virtual Earth satellite for reference. I wasn't aware thought that GE caused some sort of offset, i'll keep it in mind for my next project(s), thanks.


Indeed an excellent tutorial by Tiberius, have been using it all along as a base for the development of Lampedusa.
 
Last edited:
Top