Skip to main content
  • EN

I am trying to find a way to simplify/generalize the dataset below. There are a number of neighbouring plots similar to the one marked in blue. I would like to merge (dissolve) as many as possible of these whilst retaining the overall shape and those plots with more distinctive shapes.

The original features are lines, but in the stage shown below I have created polygons and added a centre point.

The plots I want to merge are rectangle shaped and of similar size. They are of course not true rectangles, that would be to easy, most have around 7-12 vertices.

I tried the FuzzyParallelLineFinder but inner and outer lines where all computed. I want to keep the outer. I have the feeling that dissolving polygons will give me the result I want without having to patch everything up towards the end. But I could of course be wrong about that.

If there is a easier way to do this than the one I suggest below, please let me know!

My thinking is that in my "rectangles" the distance from the centre point C to corner A would be similar to that from C to corner B. Same for the other corners. Also the angle a would be similar to the angle b. I use the word similar because I would need to set some sort of tolerance for it to work.

Here is where I get stuck. How do I extract the distances from C to A,B, D and E, and the corresponding angles? Once that is extracted, what transformer would you recomend I use to do the comparative calculations?

 

overviewcloseupp_2

Off the top of my head, one possibility could be to first create an oriented bounding box around each polygon, then use the Clipper to get the difference between the original polygon and the oriented bounding box. The remaining area as a percentage of the area before the clipping might be an indication of how rectangular the polygon is.

I like @david_r​ 's suggestion. An alternative:

  1. Generalize them to get rid of as many of the extra points as possible. Ideally you'd end up with 4 points per polygon, so true rectangles (which by themselves would be an approximation of the actual shape due to the generalizing)
  2. Create circles around those 4 points
  3. Clip the circles with the (generalized) rectangle
  4. Calculate the area of the remaining circle segments and compare them. The closer they are to eachother the closer the angles at those points are.

Off the top of my head, one possibility could be to first create an oriented bounding box around each polygon, then use the Clipper to get the difference between the original polygon and the oriented bounding box. The remaining area as a percentage of the area before the clipping might be an indication of how rectangular the polygon is.

This is what I came here to say, though I would probably use an AreaOnAreaOverlayer rather than the clipper.

 

In either case you want a unique ID on the original features and use that as a group-by.

Reply


Community Stats

30,186
Posts
114,578
Replies
37,853
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.