Salut Philippe,I mentioned something along these lines in the document concerning the use of the Autogen Annotator:Please note that excessively complex polygons may not give the desired results. This is because Flight Simulator samples the polygon to default blocks of 16 x 16 pixels. You can get greater precision by adding a key to the [TERRAIN] section of the fsx.cfg file: IMAGE_PIXELS_FOR_AUTOGEN_POLYGONS=16where 16 is the default value. By changing this to 1024, you will force the vegetation polygon to a precision of about 1 meter. Changing the value to 2048 will result in precision of about 0.5 meters. However, changing this value is not recommended as the greater precision will require more memory and processor resources.Perhaps this has something to do with what you see.Best regards.Luis

Salut LuisI think you are right. But, even if it's possible to change the precision in Fsx, I don't think it's a good idea : I want to keep the best possible Fps. I'm working on a filter able to reduce the 'star effect' and the number of points with keeping the average shape of the poly. At the same time the filter can/will reduce a bit the poly size and by this way avoid trees growing where they dont have to do. The target is mainly to produce large scale "forest" and it'll better to forget some trees than putting them on the roads or into the rivers.I have to thank you for the explanation about the 'square division' to get the name of .agn. I wrote a small function which do that perfectly : another issue fixed.Merci Luis, et bon week-endPhil

Hi Philippe:Your project seems to be showing very promising results so far; thanks for keeping us posted on how things develop. :( Luis:Very interesting info you posted (again); I have much yet to learn about such esoteric aspects of the FS Autogen sub-system... thanks again ! :( GaryGB

Don't worry, Gary, I'm still planting trees...With all the informations you gave me, almost all the problems "Fsx side" are under control. The out-files are corrects, including the strange byte swap of the Guid code. I've turned around the structure of the .bgl (the input file) with a clever remote use of TmfViewer.A 10*10 lod13 grid square (CNY_South9West23_FSXSX.bgl from Cedar Breaks National Monument http://www.blueskyscenery.com/CNY_South2West5.html ) is treated in +/- 4 minutes in a full automatic way.The last (I hope) difficulty is to explain to my stupid algorithm the difference between a tree and a farm field. For the moment, I only use luminance and chrominance thresholds, and I have to add a "shape" analysis. It's what I working now.The left side of the picture is not bad at all, even if there is some small polygons on the roads, but the right side is a mess... It's 4am now and I have to sleep a bit. Tomorow will be hard... Good night, have nice dreams :( Phil

Hi Philippe:Looks like you are using a very comprehensive approach to analysis of the source imagery, I'm confident this will continue to evolve into an even more sophisticated autogen annotation utility. :( When you have had time to contemplate more potential uses for your intermediate work files and output files, I'm curious if you might consider other uses of your analysis engine for generation of water masks and water classes ? :( Additionally, I see a potential there for also generating XML placement information for some "non-autogen annotation" scenery content (library objects, effects etc.).Furthermore, I can also see a potential for generating of XYZ coordinates for other FSX SDK CMX vector-type objects such as excludes.Possibly vector rivers or streams, and roads or other paved areas could be generated too, which can be used with default land class (and with custom land class type scenery when blend masks are used during compilation).Perhaps in time, this type of engine could even assist in automating generation of the M-Tiles required for some types of layered custom land class scenery which needs to be blended / dithered.It's rather exciting to see this remarkable utility in development; thank you for letting us all look in... as you innovate this functionality for FSX scenery work ! :( GaryGB

Wow! It's good to see what progress you make!I think with some kind of shape analysis - as you suggested - your program will fly.A functioning AutoTree tool for FSX would be very cool.Thank you very much for your efforts! 4AM :(

At last, something correct.(of course the 100 grids were treated in a batch and without any specific adjustment)I have reworked the 'trees engine' by adding a spot shape treatment limiting the acceptable trees to a small circular spot (this avoid farm fields to become trees fields), and I have smoothed the polygons filter. My current worry is to make more supple 2 or 3 parameters (of about 10) which are bit too much sensitive.Yes Gary, there is a lot of possible things to do with brother-algorithms of this one, at 3 conditions: a very good know how with the Fsx files formats (I'm a rocky with this matter), a well defined problem and the acceptance of a 95% solution. These kinds of treatment use a lot of filters (high pass which adds noise, and low pass which smooths everything) and the result is always a compromise, sometime a good one, sometime not.Two eyes and a brain is a much more powerful machine than any computer, and for a long time.Water masks, rivers, roads look possible to be produced by a program like mine. For more complex things, surely, it could help. If you have a well defined problem needing image treatment, you may email me. I'll do my best to answer.If i succeed with this program, i intend to publish the .agns as free complement for the BlueSky's sceneries, and after adding a lot of comments in English, to give the source code as a freeware, for any one to play with.Well, back to programming..Good weekend.Phil

Hi Philippe:Very impressive progress; and wonderful news too about your plans for FS Community access to the AGN annotations ....and your new utility itself ! :( I'm sure flying in Gottfried's BlueSky photoreal scenery areas will be an even more marvelous experience... thanks to your generous creations.Please be assured the FS Community greatly appreciates your remarkably innovative efforts in this area of FS Scenery Utility development. :(Perhaps as I get my pending FS projects caught up, I might be able to beta test your utility with some photoreal areas I plan to annotate. :( Keep up the excellent work ! :( GaryGB

Don

Hi philippe, Just a word from France ;-) to tell you that your new program looks very promising, the "autotrees" side seems to run with high precision. I began to annotate the photoreal sceneries that we build whith FSET in our VA for flying in aquitaine and dordogne, and no need to say that your upcoming wonderful tool may make it MUCH MUCH easier ..... Keep the good work. Jerome

Hello:IIUC, this post refers to results with tests using the utility being developed by Philippe for autogen annotation of photoreal imagery ? :( If so, I am encouraged to hear that it is still progressing, since it has been a while since Philippe posted any further news on his already impressive endeavors.I found a few links on Google referring to work in France with FSET (FS Earth Tiles) imagery and a regional France imagery source.I'm wondering if the work by Philippe in this forum thread at AVSIM is a subsequent work by the author of FSET (which was ultimately released to the public domain). or if it is just a coincidence that imagery tiles derived from FSET are being annotated by the AGN utility being discussed in this AVSIM thread ?Here's hoping all is going well with Philippe and his latest project ! :( GaryGB

Hello:IIUC, this post refers to results with tests using the utility being developed by Philippe for autogen annotation of photoreal imagery ? :( If so, I am encouraged to hear that it is still progressing, since it has been a while since Philippe posted any further news on his already impressive endeavors.I found a few links on Google referring to work in France with FSET (FS Earth Tiles) imagery and a regional France imagery source.I'm wondering if the work by Philippe in this forum thread at AVSIM is a subsequent work by the author of FSET (which was ultimately released to the public domain). or if it is just a coincidence that imagery tiles derived from FSET are being annotated by the AGN utility being discussed in this AVSIM thread ?Here's hoping all is going well with Philippe and his latest project ! :( GaryGB

No I'm not a beta tester, but 'd love to do some tests with that utility !!! I don't think there is any link between FSET and Philippe 's software.In france we make a big use of FSET, moreover Google support works again and is in some places far better than our local photos satellites provider, IGN (images from IGN can be obtained via the Virtual Earth service in FSET). Just a pic : RegardsJerome

Well this turned out to be a disappointing thread.