Military AI Works

I've tried, over the years, to make a few sceneries, modifying AFCADs to match GE images. Very often I've found that the stock airfield in no way matches the image I get from GE. Newer AFCAD-type programmes allow adding a background image to help with the design (I'm using ADE btw).

I understand how the images are taken by the satellites and therefore it's entirely possible that some distortion is bound to creep in. But how much? Does distortion occur when the image is saved and applied inside the CAD programme? And how much can we trust the GE image over the airfield which MS has given us?

I'm a little perplexed, because (taking Scott as an example) the runways and primary (parallel) taxiways line up with no problems, but all the other taxiways and parking areas are miles out.

I'm wondering, has anyone else noticed this?

I used to think that Google Earth was the "be all and end all" for accuracy in afcad creation. Then, with experience, confusion, consternation and finally some dissatisfaction set in. At this point, it's "resignation".

What I mean is this. You have to realize that freeware satellite imagery is dated to varying extents in different areas and by various sources. The primary/most often used freeware sources are Google Earth, Yahoo Satellite, MS Visual Earth and AskSatellite. These four are readily accessible via the web and Sbuilder/SBuilderX.

As I have begun to use SBuilderX more and more frequently, it's become quite obvious that I have to take each of them with grains of salt for the following reasons.
1. Lack of dating
2. Varying quality
2a. Images are assembled by splicing rectangles together. This is an automated process and errors in matching do occur - as seen when 2 sections of a runway or taxiway don't match up perfectly.
2b. Weather can and does obscure ground features as well as creating inconsistent coloring.
3. Image inconsistencies over an area. Seasonality or seasonal variances are sometimes obvious - check out Google Earth's coverage of Vladivostok.. then slide NW and check out what happens. Another area to check out is Polyarnyy and Severomorsk - green areas become sepia toned - obviously not from the same set of imagery.
4. Apparently deliberately introduced blurriness or manual distortion over an area. This is, I believe in response to government requests/restrictions relating to security concerns especially around airbases and other military facilities. Check out Leeuwarden, Netherlands.

So, where does that leave afcad or scenery creators? Sigh. With a tougher job than it first appears. What I think is the most accurate, for me, is to simultaneously slave MSFSX to AFX and to Google Earth whenever possible using FSEarth software. After drawing my afcad, I go to the same location in SBuilderX and check out the accuracy against Yahoo Satellite, MS Visual Earth and AskSatellite by slweing to various points in the afcad and using the Display Aircraft feature overlaid against the Display Background function. Usually, any disagreements are pretty obvious and can be traced to facilitiy upgrades and "imagery dating". So I tend to go with the majority view - accuracy to within 15 feet works for me 99.5% of the time.

CelticWarrior wrote:Newer AFCAD-type programmes allow adding a background image to help with the design (I'm using ADE btw).

I understand how the images are taken by the satellites and therefore it's entirely possible that some distortion is bound to creep in. But how much? Does distortion occur when the image is saved and applied inside the CAD programme? And how much can we trust the GE image over the airfield which MS has given us?

Hmm. You want specifics? Okay, try this.

The accuracy of using background images in any afcad creation program is related to "calibrating" the image to the afcad design area/window and to image resolution.

Image calibration in AFX, for example, relies on accurately locating the four corners of a rectangle-shaped picture whose orientation is along the 0-180 degree line. The more accurately those four points can be pinpointed and entered into the AFX program, the more accurately the background picture can be drawn.

Edited to add: If you "misplace" corners, the image gets stretched - normally along the north-south or east-west axis.
end edit

That only deals with part of the problem, however.

The second part of the problem with using background imagery to draw/place afcad features relates to the resolution of the picture itself. Like Xerox pictures blown up to huge sizes, inaccuracies and distortions begin to creep in the more the image is blown up in relation to its original size.

So, I can place a picture very accurately and still have a blurry image when I zoom in - something that actually prevents me from determining to a certain degree whether an apron skirt is 5 meters wide or 7.5 meters wide, for example.

I could write more... I go by the axiom, Trust but Verify, and even then I'm usually accepting a certain level of inaccuracy in what I've created.

Ford
Edited to try to correct some spelling/typo errors.

Just my take on what you are asking.

First I would trust the GE image more so then the MS version. I am sure MS did not take the time to look at overhead images for each and every airfield placed. More likely airfield diagrams, and then good guesstimates from them of the taxiways, and ramps.

On the GE images there will always be some distortion just from the fact that the pic was probably not taken from directly overhead. That being said MSFS airfields have to be larger then their real counterparts taking up more real estate then in the real world. This becomes more noticable in MSFS, when you have a number of large airfields in closer proximaty to each other. You then add more distoration when you stretch the GE photo, to get the runway to match the one in MSFS. You will also note that the ramp areas also have to be much larger then you see in GE, if you are going to try to get somewhat close to the real world parking available. This will put the joining taxiways out of kilter. So probably, we will always be in a compromise between what is actually there, and what you can model.

I think that the scenery designers have done a great job in filling out the airfields of their choice, and giving them more of that working airfield look, that by just sheer magnitude, the designers at MS could not hope to achieve.

Thanks for your replies, what you both say confirms what I was beginning to come to understand.

As with all thing to do with FS, there has to be a compromise, and that compromise is obviously accuracy. When it comes to AFCAD, specifically for AI, we have to create enough space to give room for parking. This is how I came to travel along this particular path; having created the spaces the KC-135s were all bunched up with wingtips overlapping. In this case also I consulted the airfield chart and this is significantly different to the MS offering in terms of placement of taxiways.

So I'm sticking with GE in this example and hopefully I'll have something to show for it by the weekend.

Thanks again for helping me confirm what I thought to be the case

Like Rick is saying, it comes down to compromise. I too am frustrated when I make a ramp area as accurately as possible according to GE only to find out that for whatever reason, I just can't fit the same amount of parking in that area as the overhead image shows.

You just have to do the best you can and know that there will always be one or two people who do not like your work.

C'est la vie !

MIKE JG wrote:... there will always be one or two people who do not like your work.

Have to agree with you there, Mike. It's not something I'm going to lose sleep over

The funny thing is that those people are too lazy to try the work themselves yet they don't hesitate to point out errors or things they feel are not correct.

Screw em'!

MIKE JG wrote:...yet they don't hesitate to point out errors or things they feel are not correct.

My response to such comments is pretty simple - it's easy to be critical when one doesn't understand the facts of any situation.

Part of the reason I tend to "share" with so few people/out of the public eye is that I know that those with whom I do share respect the technical complexities and occasional "difficulty" in making choices. Generally, if they don't understand the results, they ask me what references I used or considered, and possibly other questions, rather than simply publicly posting negative comments.

I have beta-tested a number of aircraft, sceneries, etc. but, in doing so, amassed a list of creators/designers for whom I no longer beta-test due to their responses to honest, objective critiques and/or their choice preferences differing from mine. This very much relates to the comment I quoted above if you think about it.

Edited to add: For me, the most accurate single technique seems to be the FSX-AFX-FSEarth/GoogleEarth combo. Since AFX doesn't accept Geotiff imagery files which have embedded geographic coordinates, I think I get better results allowing the computer programs to exchange information rather than relying on my hand-eye-mouse coordination and cursor interpolation. But if I "combine all the tools", I get something I am fairly confident can be accurate down to less than 15 feet - and I make as many parking spots for as many aircraft as can be seen in any satellite photo. For instance, I have an extended-PANC afcad (including the all-but-co-located water airport ) that can easily park over 250 aircraft without choking. OTOH, my system would choke FPS-wise if there were that many aircraft actually parked or moving on the afcad. Drawing accurate parking is rarely difficult if one is willing to take the time to use the available tools. Making traffic flow either accurately or smoothly on the afcad is something completely different.

There will always be situations that can't be solved in an afcad. Indeed it's always a compromise and it doesn't bother me that much because it's a simualtion after all.

For example I am not often very satisfied with the taxi- to taxiway or taxi- to runway connections when surfaces are different (asphalt to concrete or so). It's not easy to get that right and sometimes FS9 behaves unpredictable. I've seen asphalt surfaces even between two directly connected nodes (1000ft apart) while the surface is absolutely set to concrete in the afcad!

Then there are runways that are part asphalt and part concrete or macadam. No way to fix that as far as I know, except for creating custom scenery.

And how about runway markings that are different for both ends? Not possible.

One thing that really annoys me though: when you connect a taxiway to an apron/parking link there's often some taxiway texture missing near the apron (see image). In AFCAD you can see that, but in AFX you can't. It takes some time to tune or find a workaround like an apron underneath the taxiway end. How do you guys cope with that?



Anyway: it's almost always satisfying at the end, real or not

David

As my reply to that comment is definitely not related to the original question, I have begun a new topic.

Ford Friendly wrote: 4. Apparently deliberately introduced blurriness or manual distortion over an area. This is, I believe in response to government requests/restrictions relating to security concerns especially around airbases and other military facilities. Check out Leeuwarden, Netherlands.

You may also compare GE and VE while looking at Italian Air Bases.

I just read an article about this. While other countries expressed concerns the German Bundeswehr is "just surveilling" this.

And as for outdated images: my five-year-old house is not displayed yet. The empty acre can be seen instead. (N52° 09' 49.50" E10° 08' 56.00" if you'd like to check )

ricktk wrote:
First I would trust the GE image more so then the MS version. I am sure MS did not take the time to look at overhead images for each and every airfield placed. More likely airfield diagrams, and then good guesstimates from them of the taxiways, and ramps.

I just got into doing airfields myself recently,and from the first one I upgraded,it definitly looks like MS reference airfield diagrams from pilot guide books.

I was able to get hold of an Australain ERSA ( En Route Supplement Australia)earlier in the year,which is a book Aussie pilots carry with them which lists most of the airfields here with info in it to help them.

Two things I noticed when I did my first airfield,and from flipping through the guide and referencing what airfields MS had/hadn't done for FS2004:

1- I upgraded RAAF base Curtin YCIN,and the stock layout, including where the buildings were located, matched exactly what the ERSA reference airfield diagram showed,including the slightly less than accurate taxiways under the fighter open ended hangars(they were sort of rough drawn loops instead of accurately layed out).

This also was noticable with other airfields I checked out as well.

2- Only airfields with an airfield diagram that had runways lengths on them were included in FS.Any ERSA entries that did not have a minimum of this weren't included.

This also applied to some medium sized GA airfields in or near capital cities. Tyabb YTYA is just on the outskirts of Melbourne Victoria and was not included due to a basic text only entry in ERSA from the looks of it,but is a pretty well known airfield here and has quite a few aircraft there.

As you mentioned, MS no doubt wouldn't have the time to get detailed info in every airfield they are putting in the sim,otherwise we wouldn't get as many as they do include.

I just installed the Google Earth 4.3.7284.3916 (beta) software and it seems that evaluating the timeliness of coverage has been made much easier.

This version has a "date stamp" down in the status bar that DOES CHANGE dependent upon th elocation of the mouse/cursor.

Could both be very useful and disappointing - depending upon whether you want up-to-date-as-of-yesterday pics or just a pic.