Skip to main content
  • EN

Hello!

I have polygon geometry representing water areas obtained from satellite imagery. They are sometimes very large and include multiple rivers inside the same polygon. I would need to decompose/split those huge polygons into smaller ones in order to have one distinct polygon for each river. This is just one polygon:

I would need to identify the main tributaries and seperate them into distinct polygons. Tributaries would be cut from the main stem more or less according to the following dividing line:

I am still not sure how this could be automated given the very irregular nature of the polygon. I would also need to decide what a relevant tributary is (width of the junction and/or area upstream,...).

I have made some attempts at obtaining a sort of skeleton using the CenterLineReplacer transformer but it ends up with a crash even after generalization. I am also looking at whether it would be possible to have FME calculate the biggest inscribed convex polygons and use the result of these in order to derive a sort of dividing line between the major waterways. I did not find a transformer for that but if you think you see an approach for solving the problem, I would be very glad if you could share it with me.

I have attached the original polygon () in case you want to play around with it.

Thanks!

Olivier

I personally think you are better off drawing the division lines manually and using them in the clipping. Sometimes it's the most effiecient way....

Thank you @itay!

This is indeed something I am considering doing. But I first wanted to rule out all efficient automated options because we may need to perform these tasks for very large areas around the world. These shapes are generated from a machine learning methodology.

Thank you @itay!

This is indeed something I am considering doing. But I first wanted to rule out all efficient automated options because we may need to perform these tasks for very large areas around the world. These shapes are generated from a machine learning methodology.

So teach it to draw it right :)

 

 

Thank you @itay!

This is indeed something I am considering doing. But I first wanted to rule out all efficient automated options because we may need to perform these tasks for very large areas around the world. These shapes are generated from a machine learning methodology.

Hi @olivier,

 

even if there is an automated option probably a manual step to verify the automated step will be necessary. Tha is my experience with this kind of solutions.

 

 

Results from playing with the generalizer

 

 

 

Reply


Community Stats

30,126
Posts
114,374
Replies
37,765
Members
Terms & ConditionsCookie settings
A product of
Safe Software respectfully acknowledges that we live, learn and work on the traditional and unceded territories of the Kwantlen, Katzie, and Semiahmoo First Nations.