Sign in to follow this  
Guest adriano

Lakes, contour lines and flying objects

Recommended Posts

Help AVSIM continue to serve you!
Please donate today!

Hi Adriano,That floating GMax "plain" is a neat idea; I remember Joe using something similar a while back but don't remember whether his version was actually movable.While I think that these lake-by-lake approaches are useful for smaller projects (and a tool like Ground2K4 does provide the most control over the shape, elevation, and shoreline of water bodies), I guess that new methods capable of providing "bulk" output will be the better choice. Also, as your observation about the (lack of) SRTM quality suggests, a method that is more or less independent of the mesh would make things easier (actually, SRTM-based add-on mesh isn't all that bad: my B.C./Alberta scenery enhancements, with their new and edited lakes, are almost all based on SRTM LOD9 mesh and the water bodies fit nicely with the terrain).Chris' AutoAsm and Jim's Slartibartfast are, I believe, the most promising approaches for "auto-generating" lakes, rivers, and oceans. With either tool, hundreds of accurately placed water bodies can be generated in a few minutes. The main "chore", in this case, is the preparation of suitable water masks and, where necessary, elevation data.With the availability of free satellite images - real-color or near-infrared - making a water mask for large areas in Photoshop is a matter of a few hours at most. Even if shadows have to be edited out by hand or rivers painted in separately, which depends on the project landscape, the process is much much faster than the "click fest" in Ground2K4.Currently, the basic obstacle is the lack of freeware tools to transform UTM projections (the common data format) into the FS-native lat/long. The good news is that Jim is working on that! In the meantime, everyone looking for images as background bitmaps or water masks (e.g., for use with Autoasm, which is already available) should check out the two sites that offer satellite images already in lat/long format (see details above):--> 30-m false-color image of (almost) the entire planethttp://www.earthetc.com/ecwearth/asps/ecwe...orld/landsat742--> 10-m greyscale images of parts of the US, Europe, and Asia (select "NGA coverages", then "DOI 10m", then "Update Map")http://geoengine.nima.mil/Finally, the remaining challenge is to make the newly generated water bodies flat and at the correct elevation. Thus far, I have hand-edited the LWM source code (generated by either utility) for lakes that require flattening. However, both AutoAsm and Slarti technically allow the input of lake elevation data as secondary information; it's just a matter of finding the most efficient method for generating the elevation input. Cheers, Holger

Share this post


Link to post
Share on other sites

Hi all,Just a bit more info on 'Slartibartfast'. After a few false starts this week (three new versions sent out to testers!) I think the UTM remapping stuff is almost complete. I'm still having trouble with getting a suitable altitude for lakes, but as a workaround it can also produce a hybrid mode where sealevel is flattened but lakes (i.e. non-sealevel) are mesh-clinging. The Rift Valley area already available here is one demo; another is the Falkland Islands, currently only available on my website but which was updated yesterday, for anyone who's downloaded it previously. I'll upload this to AVSim next week if there are no problems with it. This has been produced with the latest techniques and I think it's a good example of what it can do - check out the boggy areas at the southern end of West Falkland. Some fairly large areas of the mesh have also been patched - I think it's fairly seamless even in LWMViewer.Techy details: the Falklands is made up of two adjacent SRTM tiles and three overlapping Landsat images. These have been manually edited to provide the watermask. The airports and navaids have been patched to match the new terrain. This is all internal to 'Slarti' apart from the image editing. The landclass however is from a separate database.The problem with specific lake altitudes is that it has to differentiate between lakes, rivers and larger areas which might extend off the side of the work area. How does it know whether a feature is a lake (and therefore should be all the same altitude) or a river (in which case it might slope) or a lake (even) passing into a river (sloping) ?[A HREF=http://www.jimkeir.co.uk/FlightSim/Autogen/Falklands.zip]Falklands [/A][A HREF=http://library.avsim.net/sendfile.php?DLID=47931]Rift Valley[/A]I know I've been battering on about this tool for ages now, but I want to get it right *before* putting it on general release. It's definitely being actively developed!Cheers,Jim

Share this post


Link to post
Share on other sites

Hi Jim.There's a real need for the UTM remapping. Can't wait to try it.As Holger writes, the 10meter DOI is really nice for placement, and is already in geographic projection. But, there are holes in the data. I found the Mappoint website's maps are very good for area like Switzerland, where the 10 meter images are unavailable. When I reproject them to geographic, the roads, rivers, lakes... all fall right where they are supposed to be. It's the reprojection of these types of maps and images that is stumbling budding terrain designers.Dick

Share this post


Link to post
Share on other sites

Hi,About UTM to Geographic WGS84 - I implemented a "vector" version of this conversion in SBuilder. My idea was to follow on with the possibility to reproject "raster" images as well, but I stoped that part. Now I know that Jim will provide this tool :-). If anyone want to test the "vector" part, you can find it in version 1.06. There is a datum.txt with some 18 lines for conversions between 18 datums that I use most. You can add any datum by just adding a new line with 9 parameters (the 2 ellipsoide axes and the "seven" conversion parameters).I wonder if Jim will implement a "lat/lon" export function in LWMViewer! Regards, Luisps: just one more comment about Sbuilder - it automatically generates a landclass BGL from a bitmap whose colors define the "use of land". I just discovered yesterday a site with a global worldwide land use map. I downloaded the Europe map and I will upload a SBuilder project with the map callibrated and with a simple definition of landclasses. Sbuilder now communicates with PolyView which helps a lot in selecting the Landclass textures.

Share this post


Link to post
Share on other sites

Hi all,wow, exciting developments indeed, at least for those "lower classes" that don't have commercial GIS packages at their disposal ;-)Luis, if you're talking about the GLC2000 data: good luck! They look really nice on first sight but are, in fact, an ecosystem-based classification and thus biased against diversity of human development. For example, cities, towns, industrial areas, etc., all have the same class (and, if I remember correctly, the same is true for agriculture). Similar problems with glaciers and snow (FS "needs" a buffer of "142" around "122" and a buffer of "130" around both, to avoid stange blending effects). In other words, a direct translation of GLC2000 to FS landclass makes for a pretty strange-looking and boring landscape. It would take a lot of hand-editing or clever algorithms to come up with a decent landclass, in my experience, at least for developed areas (it worked fairly well for Jim in the Rift Valley and me in Namibia).Cheers, Holger

Share this post


Link to post
Share on other sites

Hi Holger,Yes, I was referring to that database. I agree with you. I have a similar bitmap for my country which I got from another source. It has a smaller number of classes (12 I suppose) but the metro areas are much more detailed. In any case I just made a post out of it in order to show how SBuilder can be used for landclass work:http://forums.avsim.net/dcboard.php?az=sho..._id=20139&page=Regards, Luis

Share this post


Link to post
Share on other sites

Hi Holger,I'm happy to hear that in other parts of the world lakes "fit nicely into terrain". I can't say that I checked every corner of the Alps, but generally speaking I wouldn't say the same here. Of course Ground2k4 is an invaluable tool, but my results are that, either one fits the lake into terrain and gets a quite distorted lake shape, or one forces the correct lake shape and obtains a distorted terrain, i.e. cliffs and dams where they shouldn't be. I'll make some more experiments.My remark about mesh quality may be made even more general, i.e. I realize that every time I fly over areas where I've neer been in reality, they look pretty good and realistic, but when I fly over areas I do know well enough, I get disappointed and largely miss any feeling of "being there".Of course this is not only a problem of mesh and lakes, but mesh and lakes are no doubt a part of it, specially in rugged areas (which are the ones I love most). So, while I appreciate your effort towards a "bulk solution" of the missing lakes problem, and believe that it will greatly enhance the appearance of FS, my approach is at the same time more reductive and more demanding: I'm trying to understand if, maybe at the price of a lot of work, but not of disk space and frame rate, is it possible to represent the landscape, at least at some spots, with an accuracy that is only limited by the intrinsic level of detail of the scene.Maybe the analysis of the mesh quality level (and here I'm referring also to the parallel topic Steve Greenwood is carrying on in these days) should start from a comparison of the contour lines with reality. Topographers used them for centuries, and there may be a reason for it. I'm convinced that, if the contour lines of a mountain match reality closely enough, then the mountain will look realistic, and vice-versa. The only exception being almost vertical or even overhanging cliffs, which in my opinion can only be re-created as scenery objects. I've made some experiment in this direction and I think that satisfactory results are possible, but in order to fit them into terrain one must acquire total control over the background mesh. Cheers,

Share this post


Link to post
Share on other sites

Adriano, my experience with LWM lakes made with Ground2k and meshes available for the given area is that you have always to tweak manually the water mask to fit it into the mesh. This means jumping back and forth from Ground2k and the sim, taking screenshots of the area and then modifying the LWM step by step. It takes time: I've made in this way the Garda Lake, it fits quite well with the Lago mesh by Thierry Pignot.At this moment, a Ground2k scenery is largely a work of craft rather than a technical enterprise - much like embroidering or tailoring a suit!GiorgioPS: You will have surely realized that LWM masks made for a given mesh DO NOT usually fit perfectly into another mesh, even if made from the same data. As every mesh creator uses different interpolation parameters, results are different.

Share this post


Link to post
Share on other sites

Hi Adriano.The problem is not only with SRTM mesh, but with all mesh.The sampling of elevation has been done by either survey or ( now ) by satellite radar reflection. Survey is most probably the more accurate, by far. New surveys make use of GPS for elevation, and should be extremely accurate.In addition the SRTM is raw data, and still needs mathematical "coaxing". I don't have the specs for the mission right with me as I write, but the released 3 arc second data is probably accurate to about + or - 75 meters for Lat/Long and + or - 15 meters for elevation. That's still quite a span for error, especially concerning the shape of lakes. Remember, the 3 arc second data is derived from the original 1 arc second data ( that will never be released to the public ). The more the data is "resampled" by NASA, the poorer it's quality will become for endusers.As I wrote above, there are some sources of maps that can be reprojected to geographic ( Lat/Long ) and they are pretty accurate as to the shape and location of lakes, roads and towns. I think I mentioned MapPoint. The 10 meter DOI is also very good, but not available for much of Switzerland.In the US, we can use TIGER data to derive images for individual counties ( my county is about 25km x 20km ). This data is survey data, and is great for lakes, roads, streets, etc. I know European countries also have survey data of very good accuracy, but it's expense is terrible.AutoASM was being developed to take advantage of maps so the drawing of Ground2K4 could be bypassed, and the image itself used to provide an accurate shape. I don't know the current status or availability of AutoASM. It looks as if Jim Keir may also add something similar to his new program, so lake shapes can be superimposed onto the terrain.==================================We'll never get accurate lakes or streams from elevational data, as the data is close, but not exact. Maps derived from surveys will be the most accurate, followed by orthographic photos ( either aerial or satellite ). Photos must be mathematically adjusted for camera angle, and for the lens distortion near the edges, and for the curvature of the earth. For example, the DOI 10 meter images have already had this done.In the US we have NED elevational data that accounts for most lake and river elevations ( the data has been adjusted using survey data for lake shapes and elevation ). Even when using this superior NED data, lakes in FS9 need elevation for the LWM poly, as resampling the mesh data "shrinks" the lakes slightly, and a mesh-clinging water poly will crawl up the banks of the lakes and rivers. Fortunatly, for NED derived mesh, we can get the elevation right from the mesh, so everything fits rather well.If Ground2K4 has a good image, georeferenced and projected correctly, you should be able to draw a lake or river accurately. And it's elevation could be confirmed by map data or through the internet... knowing that elevations of water do change seasonally and sometimes annually. So if a good map makes the lake seem out of place in the sim, the mesh is at fault... probably because the mesh source data was wrong or too course. The only choice then is to try to adjust the surrounding mesh elevation, or adjust the lake/river elevation. As with airports, usually the surrounding mesh elevation gets adjusted... and rightly so, as the mesh is most probably wrong.Sorry for the long post. Summarized:Maps good, mesh bad.DickPost-script:This is one reason why I do not like mesh in the sim sampled to over LOD7. The data really is not accurate, no matter the source, and we make it less so by the resampling we need to make. If we are trying to get a more "bumpy" effect, then limiting the mesh to LOD8 might be fine. Over that level, we get mesh "popping" and we are also overtaxing the CPU... CPU time that might be better spent accounting for 3d clouds or vector-autogened telephone poles. ;)Roads, rivers, lakes and city shapes do add a greater sense of familiarity to the sim, and I would give them more preference than the LOD level of the mesh. And landclass can also be a treat for a home area when it is well done.

Share this post


Link to post
Share on other sites

Hi Dick,I've been making some further experiments with meshes - freeware meshes of the Alps, SRTM and not - and I totally agree with your statement: mesh is bad. I'd like to set up a method to check meshes by comparing their contour lines with maps, anyway my first results are astonishingly disappointing. It just seems that correspondence with reality has never been considered as an important factor for mesh quality. So far, Giorgio is right as well: today, re-shaping lake shorelines in rough terrain, or creating a new lake, appears closer to an art (the art of tradeoff!) than to a scientific method.Yet, I think, it shouldn't be. The conclusion may be that any mountain scenery claiming to be realistic should be backed by a small, accurate mesh built to the purpose. But the basic question still remains open, at least to me. Let us suppose that one makes a nice job with LC and VTPs, maybe special textures. In a rugged area, where relief modelling is vital for landscape appearance, is it actually possible to obtain a mesh that closely matches reality within the current Fs9 limitations?On the other hand, I think you are overly pessimistic in proposing to ban any mesh beyond LOD7. At least in places where I've never been, LOD9 or LOD10 meshes do enhance my pleasure of simming. Of course I'm always talking of mountain areas. Maybe we should strive for a simulator that LOOKS realistic everywhere (and LOD7 doesn't usually convey that feeling), and IS realistic where someone has devoted a special care to make it so. Again, provided that this is possible.----P.S. Don't know if this may be helpful, if not please disregard. I remember that time ago you were investigating a small disalignment between LWMs and VTPs. In these days, I realized that Ground2k4-created coastlines are actually displayed by TMFviewer if you first load a background mesh. Well, what I saw was that the associated LOD13 gridlines DO NOT appear exactly aligned with LOD8 gridlines: they seem to be just one LOD13 pixel shifted west and north.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this