Legacy Question: Forest Map creation from photoimagery for V1?

Moderators: Uros, OXO, JBr

Post Reply
User avatar
M Pompian
Posts: 18
Joined: Thu Jan 08, 2015 8:46 pm

Legacy Question: Forest Map creation from photoimagery for V1?

Post by M Pompian » Sat Jun 13, 2020 11:10 pm

Is such a thing possible?

Can the V1 .for (forest) file format be addressed with Science?

There must be SOME procedural method to help me finish this map...

Thanks for any help!!

M. Pompian
Image

Reg.#: N1953MP
Comp.#: AVA

User avatar
M Pompian
Posts: 18
Joined: Thu Jan 08, 2015 8:46 pm

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by M Pompian » Fri Jun 19, 2020 11:19 pm

With 64 views and no response, I have to believe the answer is no.

Still, thanks forum for getting me through this archived project.

M. Pompian
Image

Reg.#: N1953MP
Comp.#: AVA

solidxsnake
Posts: 12
Joined: Sat Apr 18, 2020 10:38 pm

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by solidxsnake » Thu Jul 02, 2020 4:21 am

In my NE landscape, I used the National Map Tree Canopy Cover raw data, followed by importing the data into QGIS and using the merging/warping tools to get the tiles to the size/resolution I needed them.

bluefang
Posts: 85
Joined: Thu Jun 11, 2015 2:55 am
Location: South East, USA

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by bluefang » Sat Jul 04, 2020 4:45 am

solidxsnake wrote:
Thu Jul 02, 2020 4:21 am
In my NE landscape, I used the National Map Tree Canopy Cover raw data, followed by importing the data into QGIS and using the merging/warping tools to get the tiles to the size/resolution I needed them.


Wow... just happened to stumble on this reply. I am currently downloading the CONUS data set - so I haven't had a chance to verify how good the data is, but if it is accurate - what a time saver.

The wand select tool in Photoshop is not at all capable of differentiating between trees, lawns, fields, farms and swamp. I wonder what they are using to gather this data? There is an error value associated with the data - I wonder if they are using ML algorithms, imagery + radar + IR processing, or brute force human processing? Probably not brute force human processing :)
Image

Get my scenery on condor.club: RDU-USA (Raleigh/Durham area of NC)

bluefang
Posts: 85
Joined: Thu Jun 11, 2015 2:55 am
Location: South East, USA

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by bluefang » Sat Jul 04, 2020 5:05 am

I might post this question in a separate thread - but seeing as how it is very much related to canopy cover / forest data...

Given we end up with a decent raster file that represents canopy cover / trees (hoping that the National TCC data set is accurate) - I am thinking about a GIMP / Photoshop plugin that takes this single set raster data - and spits out two different images - corresponding to the percentage cover of coniferous vs deciduous for a tile.

For example - lets say we process the TCC data using qGIS and end up with an appropriate forest tile (2048x2048 where color A pixels represents canopy / tree data, and color B pixels represent nothing). What I would like is a way to specify that the canopy coverage is 20% Coniferous and 80% Deciduous and have this plugin produce the appropriate sxxx and bxxx files automatically. Think of the time saving. At the moment, I haven't even begun the process of integrating the two different types of trees into my scenery, as the area of my scenery has no real delineation of the two - there is usually just a percentage of one vs the other all clumped together.

Man, who knew making landscapes was this difficult...
Image

Get my scenery on condor.club: RDU-USA (Raleigh/Durham area of NC)

User avatar
EDB
Posts: 1575
Joined: Mon Oct 31, 2011 1:17 am
Location: The Netherlands, Europe Continent, Earth Planet, Milky Way Galaxy, Virgo Super Cluster

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by EDB » Sat Jul 04, 2020 3:59 pm

bluefang wrote:
Sat Jul 04, 2020 5:05 am
Man, who knew making landscapes was this difficult...

Landscape Builders...

User avatar
dgtfer
Posts: 521
Joined: Mon Nov 19, 2007 9:49 am
Location: Marseille - france

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by dgtfer » Sun Jul 05, 2020 3:35 pm

bluefang wrote:
Sat Jul 04, 2020 5:05 am
Given we end up with a decent raster file that represents canopy cover / trees (hoping that the National TCC data set is accurate) - I am thinking about a GIMP / Photoshop plugin that takes this single set raster data - and spits out two different images - corresponding to the percentage cover of coniferous vs deciduous for a tile.

For example - lets say we process the TCC data using qGIS and end up with an appropriate forest tile (2048x2048 where color A pixels represents canopy / tree data, and color B pixels represent nothing). What I would like is a way to specify that the canopy coverage is 20% Coniferous and 80% Deciduous and have this plugin produce the appropriate sxxx and bxxx files automatically. Think of the time saving. At the moment, I haven't even begun the process of integrating the two different types of trees into my scenery, as the area of my scenery has no real delineation of the two - there is usually just a percentage of one vs the other all clumped together.
Where will you find a map with these percentages? All we can find is maps with either coniferous either broad-leaved either mixed, but with no repartition...
But if you found this information, it is "simply" doable by creating a map with layers for the various zones and applying a B&W noise filter with the proper intensity for each zone, then you must merge these layers and create the s and b images. For that, you'll just inverse B&W in the forested zones.
Image

solidxsnake
Posts: 12
Joined: Sat Apr 18, 2020 10:38 pm

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by solidxsnake » Sun Jul 05, 2020 4:07 pm

The data's quite good, but it won't have any split between coniferous vs deciduous. I wasn't worried about that personally, I made all the maps in my scenery deciduous. Remember that the maps are set up to only have one type per pixel (in the 512x512 map). If you wanted to do percent-splits of the coverage, you'd probably want to do some form of dithering.

The National Map TCC data is quite good, but there is certainly going to be some touch-up to do if you want it to be "perfect." There will be some stray trees in areas like highways/roads that you'll likely want to remove. I'm not too worried about it for the first-pass of the NE landscape I'm working on, since the primary purpose is primary training for students at my local glider club. It's not anything too noticeable at altitude, and very tedious to update, so beyond some minor touchups in the patch that contains the club's home airport, I haven't bothered doing any touch-up.

Working on making base thermal maps for the landscape today, then hopefully uploading it to Condor-Club for everyone to use. Only a couple airports have asphalt/grass put into place, with 3D objects only done at the club's home airport, but adding in airports later is easy enough (the terrain has already been flattened, so only need to add the O/G .c3d files).

bluefang
Posts: 85
Joined: Thu Jun 11, 2015 2:55 am
Location: South East, USA

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by bluefang » Sun Jul 05, 2020 9:50 pm

dgtfer wrote:
Sun Jul 05, 2020 3:35 pm
Where will you find a map with these percentages? All we can find is maps with either coniferous either broad-leaved either mixed, but with no repartition...

I don't expect to find a map specific to tree type - simply using the generic canopy cover is good enough. At least in the area for my scenery (North Carolina, USA) - we can make a "ballpark" estimate of coniferous vs deciduous density - as it is pretty consistent throughout the entire area of my scenery. The density ratio may change as you get closer to the mountains - but in my flatland landscape - the ratio is going to be static.

dgtfer wrote:
Sun Jul 05, 2020 3:35 pm
But if you found this information, it is "simply" doable by creating a map with layers for the various zones and applying a B&W noise filter with the proper intensity for each zone, then you must merge these layers and create the s and b images. For that, you'll just inverse B&W in the forested zones.

Ah - yes. I had forgotten that the Photoshop noise filter had a "percentage" slider. For some reason, I though it was always a 50% noise algorithm.

I just did a quick-n-dirty test to create data needed for the sxxxx and bxxxx files.

Here are the steps I used in Photoshop:

  1. Start with a single canopy cover layer - where canopy cover is represented by a pixel of any color, and no canopy cover is represented by transparency. This should be easy to acquire from the canopy cover data set.
  2. Duplicate this layer
  3. In one layer, add a layer effect -> color overlay - and choose something like green
  4. In the other layer, add layer effect -> color overlay - and choose something like purple
  5. Create a new layer, fill it with solid white, and then Filter -> Noise -> Add Noise - select the desired percentage of either coniferous or deciduous (doesn't matter), make sure monochrome is checked and Gaussian is checked.
There may be faster ways for the remaining steps, but this is what I am use to doing:

  1. With the noise layer selected, click on the "Channels" tab for that layer - then Ctrl + Click on the RGB channel - this will add all of the white pixels to the selection.
  2. Now on one of colored canopy layers - click on the "add layer mask" button for that layer.
  3. Unfortunately, that clears our selection, so go back and and reselect all of the white pixels again in the RGB channel of the noise layer.
  4. Select the other canopy cover layer this time, and Alt + Click on the "add layer mask" button for that layer - this will create an inverted mask


Results (this one is 80% green and 20% purple):
tcc-density-photoshop.jpg
and zoomed in a bit:
tcc-density-photoshop-zoomed.jpg

Now off to figure out if this is easier to automate / script in Photoshop or GIMP
You do not have the required permissions to view the files attached to this post.
Image

Get my scenery on condor.club: RDU-USA (Raleigh/Durham area of NC)

bluefang
Posts: 85
Joined: Thu Jun 11, 2015 2:55 am
Location: South East, USA

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by bluefang » Mon Jul 06, 2020 12:47 am

solidxsnake wrote:
Sun Jul 05, 2020 4:07 pm
The data's quite good, but it won't have any split between coniferous vs deciduous. I wasn't worried about that personally, I made all the maps in my scenery deciduous. Remember that the maps are set up to only have one type per pixel (in the 512x512 map). If you wanted to do percent-splits of the coverage, you'd probably want to do some form of dithering.

The National Map TCC data is quite good, but there is certainly going to be some touch-up to do if you want it to be "perfect." There will be some stray trees in areas like highways/roads that you'll likely want to remove. I'm not too worried about it for the first-pass of the NE landscape I'm working on, since the primary purpose is primary training for students at my local glider club. It's not anything too noticeable at altitude, and very tedious to update, so beyond some minor touchups in the patch that contains the club's home airport, I haven't bothered doing any touch-up.

I was able to get the National TCC data imported into qGIS today and I would say it is indeed a good starting point for tree cover - but there is indeed quite a bit of touchup needed. It will be great for covering large swaths of scenery where there are no airports or other areas that require hi-res / accurate tree placement - but I realized after importing, that the canopy cover data is 30m resolution - which is quite low for detailing the areas around airports. And I am seeing quite a few areas where the 2016 canopy cover doesn't match my 2016 aerial imagery - including entire farms that are shown as covered by trees as well as some lakes covered by trees.

So the way I see it is that the data is a good starting point - but to make your scenery realistic, it is still gonna require quite a bit of touchup work.

Here is an image from my scenery close to one of the local glider airports - and areas where trees shouldn't be pointed out with the arrows (several lakes are covered in trees as well as several farms):
canopy-cover-diagnostics.jpg
You do not have the required permissions to view the files attached to this post.
Image

Get my scenery on condor.club: RDU-USA (Raleigh/Durham area of NC)

User avatar
dgtfer
Posts: 521
Joined: Mon Nov 19, 2007 9:49 am
Location: Marseille - france

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by dgtfer » Mon Jul 06, 2020 11:00 am

Have you tried the USGS NLCD (National Land Cover Database)? It seems to be quite reliable, and its resolution is often better than 10m.
Image

User avatar
EDB
Posts: 1575
Joined: Mon Oct 31, 2011 1:17 am
Location: The Netherlands, Europe Continent, Earth Planet, Milky Way Galaxy, Virgo Super Cluster

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by EDB » Mon Jul 06, 2020 11:08 am

NLCD is 30m... And quiet old.

bluefang
Posts: 85
Joined: Thu Jun 11, 2015 2:55 am
Location: South East, USA

Re: Legacy Question: Forest Map creation from photoimagery for V1?

Post by bluefang » Mon Jul 13, 2020 6:38 am

I posted this in the GIMP plugin thread, as it is indeed a GIMP plugin - I created a GIMP plugin that does exactly what I had previously inquired about...

It takes a tree map input file (a 2048x2048 image file with non-transparent pixels as trees and transparent pixels as not trees - color doesn't matter).

It presents a dialog where you can input

  • Tree Density - the percentage chance that any non transparent pixel is a tree pixel. Usually want to keep this at 100.
  • Deciduous Density - the percentage chance that any tree pixel is a deciduous tree vs a coniferous tree. Any tree that is not a deciduous tree is automatically a coniferous tree as Condor doesn't seem to allow overlapping tree types.
  • You can also specify the pixel colors for each tree type in the bXXXX, sXXXX output files - but these colors don't seem to matter as long as they are different from each other.
The input file must have an alpha layer (for defining transparency) - and should be 2048x2048 - and should be named using the Condor Terragen tile naming convention, i.e. "CCRR.png". The plugin will create corresponding bXXXX.bmp and sXXXX.bmp files based on the dialog values.

Plugin can be found here:


https://github.com/johnny-fantastic/condorv2-tile

Plugin Dialog:
plugin-dialog.jpg


Example Input File:
tree-input-file.jpg


Results:
results.jpg
You do not have the required permissions to view the files attached to this post.
Image

Get my scenery on condor.club: RDU-USA (Raleigh/Durham area of NC)

Post Reply