Studying image re-projection of UTM nad83 image

[Guest]

Studying reprojection per rhumbaflappy

[Holger 0]

Howdy:haven't worked with orthoimages in FS yet but lots in my professional endavors. "Rectifying" is also known as rubbersheeting and is used to minimize distortion of imagery due to steep terrain or problematic camera geometry (camera tilt & lens distortions). To do this, you have to pick ground reference points with known coordinates, find those locations on the image, and then have the software run a polynomial to "tug and stretch" the image as required.In contrast, reprojections are predefined linear (more or less) algorithms to transfer between different projections and datums. From my work with FS add-on meshes, I noticed a systematic offset between the FS default hydrography and the add-on mesh. This could be partly due to a goof-up with datums (e.g., NAD27 instead of NAD83) but wouldn't lead to an angular distorion as in the case of your runway. More likely, it's a problem with the database FS used for making airfields (I know of several incorrect locations so incorrect runway directions wouldn't surprise me).I don't think rectifying is the best choice in your case as you don't seem to have much relief in your area and your image was probably georectified to begin with (or did you just scan a hardcopy of an airphoto?). On the other hand, you could try it if you can't get the runway to fit any other way (and if you can find easy-to-learn free software that allows it); however, it probably wouldn't be perfect and might look funny as some of the straight lines would end up being crooked.How about this: load the image into Photoshop (or something similar) and manipulate the image itself; cut out the airfield and either rotate it or copy another part of the image into the offending area. I can give you more pointers on that, if necessary.Cheers, Holger

[Guest]

Thanks for your response! And I appreciate the teaching on rectifying! The image was from microsofts terraserver, its the usgs doqq, so it may be rectified already.Funny, the same thought occured to me, to reposition the airport, somewhat artificially amoungst the rest of the photo. While that could solve the problem, I'm still interested in what the cause is. I suspected initially that the runway was not angled properly, but I went to the airport and interviewed the managers...they swear that the runway heading is really 14 magnetic..seemed to me that 4 degrees might be off without the pilots taking note, but its not the case, aparently.Best,Bob Bernstein

[nudata 10]

Bob,Regarding your observation on MS road placement. I have always judged the road placement accuracy of MS roads, rivers, etc. as their alignment with the underlying mesh. I have never seen any of the MS items come close to matching the mesh. This means, where I know from personal observation, that a major road follows a valley the MS road does not. In fact the actual shape of the road, even if moved, could never follow the valley. However, as you point out, they may be closer than we believe. This may be especially true when we use third party mesh as an addon. I have always placed my "faith" in the accuracy of the mesh. However, if it is not aligned to the Microsoft version of the Earth (whatever that is???) then we may be making a somewhat bad situation, worse. Perhaps someone who is skilled in cartography could shed some light on this - please.....Regards,Dick KLBE

[Holger 0]

Howdy:well, there are two basic problems here: one is the displacement of the FS linework (rivers, roads, shorelines) relative to their true geographic location and the other is the small (i.e. coarse) scale of the source data MS used for generating the linework.The latter explains why moving linework to match the mesh (or vice versa) will not provide a consistent improvement, particularly in mountainous areas. Since the scale of the imported linework (1:2 Mio, as far as I know) determines the spacing of points along a line, many curves and corners are shown as straight lines and thus make all sorts of strange-looking shortcuts. More accurate (large-scale) data are available for free for some areas, including North America, but it's no easy task to turn them into FS-usable .bgl files. See this thread: http://forum.fsgenesis.net/cgi-bin/dcforum...11&viewmode=allAs for the locational displacements, things are similarly challenging. As a scenery designer, you'll have to pick a particular mesh (default or add-on) to work with but that often means your scenery will conflict with the other mesh options for the same area. As anyone making or using add-on meshes with lots of lakes or coastlines knows there is a definite mismatch between these meshes and the FS shorelines. Part of this problem is due to the linework generalization described above but there seems to be an additional, more or less consistent geographic displacement as well (the image in the post cited above shows this displacement well). Since mesh designers know that their data are accurate in terms of location, ideally we'd like to remake or move the shoreline data to match better with the true position. Unfortunately, that is almost impossible to achieve, at least on a regional or continental scale.My mesh designs have used the opposite approach, that is, deliberately mispositioning the add-on mesh tiles to match with the incorrectly located FS linework. It's a bit tedious but works quite well, and I know of at least one other person (Gilles Gauthier) who has done the same. The added benefit is that this approach tends to reduce the problems of lakes wedged into mountainsides and raised or sunken lakes and airports. Unfortunately, the problem remains that scenery designed for add-on meshes that don't use these methods will probably not look so hot. For example, many of the scenery add-ons made to fit with the popular FS2000 SE Alaska mesh by Ed Denney don't fit well with my mesh for the this region. Catch-22 ;-(In summary (I guess), us designers have to choose one of the options and stick with it; there is no right or wrong, just different approaches.Hope this helps. Cheers,Holger

[Justin Toposim 16]

HI Bob,On of the problems with simply "rotating" a DOQQ is that it isn't a perfect rectangle. To get good results you have truly reproject it (or better said, unproject it, and not just rotate it. UTM projections are always wider at the bottom than at the top (at least in the northern hemisphere). If you merely rotate the image, you're still going to end up catawompus.I don't know if Microdem can take a UTM orthophoto and "unproject" it, which is what you want to do.Maybe if you have a birthday coming up, or an extra $250 laying around, and are "serious" about the use of GIS in constructing FS scenery and terrain, you might want to investigate Manifold. It's a relatively inexpensive (as GIS' go), yet powerful application that will do many helpful things for designers.http://www.manifold.netI've been using it for two or three years for various and sundry tasks, and it's probably the one program I use most in my research and development. I find it indespensible. Absolutely indispensible.-------Justinhttp://www.fsgenesis.netHigh Quality Scenery for FS200x