Integrate Planet UDM masks to filter clouds and unusable pixels.

[SamtaSakha]

Currently, the coastline is derived from the binary water mask produced by the segmentation model.

I would like to explore improvements to the raster-to-vector conversion process by:

• removing small noise polygons
• smoothing extracted coastlines
• simplifying geometry for cleaner vector outputs
  This could improve the quality of coastline datasets used for erosion modeling.

[praveen-86176]

Hi, I’d like to improve the raster-to-vector coastline extraction.
The plan is to remove small noise polygons, smooth the coastlines, and simplify the geometry.
This should result in cleaner and more accurate vector outputs for erosion modeling.
Please let me know if I can proceed with this improvement.

[modi02]

This is an interesting problem. From what I understand, after the model outputs a binary water mask, converting that to a clean vector coastline is tricky because the raw output has jagged edges and small noisy patches that don't represent real coastline features.
A few things I'm wondering about as I read through this:

Would a minimum area threshold work for removing noise polygons, or does the coastline geometry in Alaska vary too much for a fixed threshold?
For smoothing, has anyone looked at whether the smoothing should be different for cliff areas vs flat beach areas?

Still learning the codebase but wanted to share these thoughts. Happy to help explore this further.