How to get single trunks as features out of a forest (aerial image)

[totas]

Hey there,

I hope some of you can help me with my problem.

There is an aerial image of a forest where you can see the result of a so-called downburst. In a certain area trees got knocked down by the wind.

Extract:

My task is to create a program with the model builder, which uses this aerial image to give out

1. in which cardinal direction the trees on the ground are pointing

AND

2. what the diameter of each trunk is.

My biggest problem is that I don't know, how to filter out the single trunks out of a all over brown forest in the wintertime and how to get them as features whereby I can calculate the further goals with.

I'm using ArcGis 10.3.

I would be so greatful if someone would have an idea, a tip or a solution for me.

Thanks,

tt.209

[ChrisSmith7]

Hmmmm, I don't have easy answers to your questions, but... is that the only data you have with which to work? I'm going to go out on a limb (hah) and say you're going to need something more useful than an aerial image. You might want to consider lidar:

http://www.tandfonline.com/doi/abs/10.5589/m13-013

Maybe Werner Mücke has a few pointers - https://geo.tuwien.ac.at/staff/werner-muecke/publications/

[totas]

LIDAR is a great idea, but too expensive for me... 😕

[ArthurCrawford]

Hello Tobias,

Does all your imagery have snow covering it like this where you can see lines of snow on the downed trees?

Thanks,

Arthur Crawford

[ArthurCrawford]

Hi Tobias,

If the snow is present like this in the imagery, this process below does not answer your entire question.   It might give you some ideas.    I took the image and applied a filter with extract band function for the 2^nd band and then remapped it 0 – 229 as no data and 230 – 255 as 1.   I then used Raster to Polygon tool.   I ran a Minimum Bounding Geometry with Rectangle by Area.    I calculated a field called L2W with (([Shape_Length] /4) *( [Shape_Length]/4))/ [Shape_Area].    I then did a definition query of  Shape_Area < 500 AND L2W > 1.4.   I added a Long field called Angle and used the Calculate Polygon Main Angle on the rectangles. 

I think it would be really hard to extract every tree trunk from this automatically and get the diameter.   What is the resolution of your imagery?

  

Thanks,

Arthur Crawford

[ArthurCrawford]

I missed a important step in the process as I was writing it, it's below in bold. "I took the image and applied a filter with extract band function for the 2^nd band and then remapped it 0 – 229 as no data and 230 – 255 as 1.   I then used Raster to Polygon tool. I ran an Aggregate Polygons with a Distance of 3.   I ran a Minimum Bounding Geometry with Rectangle by Area.    I calculated a field called L2W with (([Shape_Length] /4) *( [Shape_Length]/4))/ [Shape_Area].    I then did a definition query of  Shape_Area < 500 AND L2W > 1.4.   I added a Long field called Angle and used the Calculate Polygon Main Angle on the rectangles."

[totas]

Great!!

Thank you very much!!

One questions regarding this:

Which scale did you choose for the distance? (Aggregate Polygons)

[ArthurCrawford]

The image was copied from above after opening and saving.   The scale was not present in the image, I used the default resolution of the image, set the distance to 3 and left the units as unknown.

Arthur Crawford

[totas]

Ah, sorry, I meant units. If I set the distance to 3 and the units as unknown it automatically sets "meters" as unit. In my case it creates huge polygons.

Which unit or which value with which unit should I choose? Whereof it depends on?

Thanks

Tobias

[ArthurCrawford]

It depends, I would start with the pixel size of the imagery time 3.   If this does not work, increase in adjustments until you get aggregated polygons that look like they could become line like polygons.   The selection process gets rid of the big polygons you see.   You may have to play with changing the setting to the definition query (Shape_Area < 500 AND L2W > 1.4).   This is what you want it to look like after the aggregate has been run.

Thanks,

Arthur Crawford