Hi all,
I’d like to classify polygons by shape, these might include: pie shaped, L shaped, T shaped.
I was originally thinking that I’d have to convert them to raster and use some sort of supervised classification. Another method might be to convert them to centerlines and compare their linework: a T shape polygon would have 1 or 2 lines at the top, meeting at 90degrees to another line point downwards
Solved
Classifying polygon by shape
+4- Contributor
Best answer by ctredinnick
For Circle, Square, you have bounding box and circularity approaches. L, or T shaped sounds like you’re interested in the number of concave parts of the shape.
You could Generalise it to remove noise/complexity, Thin, with a tolerance like half the distance across the polygon. HullReplacer, then clip it against itself to count the concave parts.
It’s difficult to know the limitations because it’s dependent on how you’ll use it.
A raster classification would be viable if one already exists. If it doesn’t then you have to weigh the time to make it, and how long it will be useful for.
+47- Evangelist
Between what kinds of polygon shapes would you have to differentiate?
Number of points can be an indicator, the angles of the polygon, the number of angles that are larger than 90 degrees, the lengths of the sides, is the polygon convex or concave, maybe the ratio between area and circumference, and much more.
+4- Contributor
Hi geomancer
We’re looking to basically classify non-square or rectangular polygons.
These would have to be classified as pie shaped, L shaped, T shaped.
I think you’re right, an analysis of points, lines and angles could be very useful.
I found that initially excluding squares and rectangles (which we don’t care about) can be done by a bounding box transformer - then comparing the areas of both. They should be very similar
Perhaps some sort of AI tool or raster classification could also work?
+51- Celebrity
The CircularityCalculator can find more circular features - https://docs.safe.com/fme/html/FME-Form-Documentation/FME-Transformers/Transformers/circularitycalculator.htm
+18- Supporter
- Best Answer
For Circle, Square, you have bounding box and circularity approaches. L, or T shaped sounds like you’re interested in the number of concave parts of the shape.
You could Generalise it to remove noise/complexity, Thin, with a tolerance like half the distance across the polygon. HullReplacer, then clip it against itself to count the concave parts.
It’s difficult to know the limitations because it’s dependent on how you’ll use it.
A raster classification would be viable if one already exists. If it doesn’t then you have to weigh the time to make it, and how long it will be useful for.
Reply
Rich Text Editor, editor1
Editor toolbars
Press ALT 0 for help
Helpful Members This Week
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.