I have parcels polygons into a databse and I want to determine what kind of geometric shape (square, triangle, rectangle, trapeze, other) each of those polygons have.
i am using the circularitycalculator and after a use a testfilter to determine each shape by the shape value (square ~ 0.75, rectangle ~ 0.65, etc).
Unfortunately, a lot of polygon have weird values, so my filter doesn't work well.
Would you have a better idea?
Thanks!!
Assume that there is no midway point on every edge of the shape.
Count number of coordinates (CoordinateCounter).
In FME, a polygon of N-vertices has N + 1 coordinates, since boundary of a polygon has end point which is on the same location as start point. Therefore:
If the number is 4, it's a triangle,
if the number is 5, it's a tetragon,
and so on.
Classify a tetragon. Square is a rectangle; rectangle is a parallelogram; parallelogram is a trapezoid; trapezoid is a tetragon.
http://en.wikipedia.org/wiki/Quadrilateral
If a pair of edges facing each other is parallel, it is a trapezoid, and if another pair is also parallel, it is a parallelogram.
For a parallelogram, calculate the angle which two adjacent edges make. If the result is right angle, it is a rectangle, and if two adjacent edges have the same length, it is a square.
I think the vector operations would be useful in determination of parallel and right angle. But you always need to be careful about a computational error.
Vector and length of edges can be calculated after transforming the tetragon into individual edges with the Chpper (Mode: By Vertex, Maximmum Vertices: 2). Or Python scripting is also possible.
Alternatively, maybe the BoundingBoxReplacer (Replace With: Oriented Bounding Box) can be also used effectively to classify a tetragon. Comparing original tetragon with oriented bounding box, you can determine whether the tetragon is a rectangle.
There is always more than one way.
Takashi
Prior to the topology detect operation i suggest to simplify the objects. As we are talking polyline type objects, it is easy to remove superfluos vertices.
You also might want to do a edgecounter prior to the search.
Would (possibly) greatly reduce search operations.
In city of Leiden (where i live) the chances of finding perfect mathematical topology is neigh zero.
have a nice Fme session.
Gio