How to detect Polygon shapes from satellite images?

Hi guys,

I am a newer one for Deep Learning and AI technologies. My mission is to find polygon shape (such as lake, bunker, hole …etc) from the satellite image if center coordinates of polygon shape are giving.

I found an example from scikit Library: https://scikit-image.org/docs/stable/auto_examples/segmentation/plot_rag_merge.html#sphx-glr-auto-examples-segmentation-plot-rag-merge-py

But no success.

Do you have any solution for my case ?

Thanks in advance.

Welcome to the community @tugsjargal.
The standard reply would be “show a sample image that you tried”… as that helps us to answer your question.

Hi @eljonco

Thank you for your immediate response. I am very glad to join to this great community.
I hope to learn more experiences.

Here is my image example.
satellite_img

Little to do with AI or Deep Learning, I would perform the following macro, which yields most lakes into a Region of Interest in the ROI manager. I only went for the lakes, but you can change the thresholding options and methods to separate by colour (Plugins> Filters> Color Deconvolutions; ‘by ROI’, then select up to 3 different specific color areas; see color deconvolution or (paywall?) this method).

Various macro commands give you access to the outline coordinates and options to convert to polygons by changing the smoothing or interpolation.

run("8-bit");
run("Invert");
setAutoThreshold("Default dark");
//run("Threshold...");
setAutoThreshold("Minimum dark");
run("Analyze Particles...", "size=100-Infinity display exclude add");

Thank you @eljonco

I will try and share my result :slight_smile: