SpatialRelator with 3D data

Hi,

 

 

I agree with Gio.

 

The SpatialRelator with "INTERSECTS" detects not only the feature sharing boundary but also the feature touched at a vertex. And also, if you have sent all the features to both the Requestor and the Supplier ports, the transformer detects the feature itself for every feature, unless you specify unique feature ID attribute to the "Attributes(s) that Must Differ" parameter.

 

Maybe it's the reason the "_related_candidates" has 3.

 

 

In addition, in my testing, the SpatialRelator (FME 2014 SP4) seems not to determine kind of spatial relationships as expected when the geometries are faces.

 

I would try this.

 

1) To avoid the wrong determination, once transform the triangular faces into 3D polygons with a GeometryCoercer (Geometry Type: fme_polygon).

 

2) Add a SpatialRelator.

 

Tests to Perform: TOUCHES

 

Calculate Cardinality of Intersections: Yes

 

3) Based on the cardinality calculation result, determine which of "line" (sharing boundary) or "point" (touched at a vertex) was the intersected situation.

 

4) After some required processings, re-transform the polygons into faces with a FaceReplacer.

 

5) Group the faces for each building, aggregate them (Aggregator).

 

 

Just be aware that the SpatialRelator (FME 2014) has a bug in the cardinality calculation for overlapping of areas, although it doesn't affect your case. The bug has been fixed in the FME 2015 beta.

 

 

Takashi

The SpatialRelator might not be a best choice for the purpose.

 

The point would be how to group consecutively adjoined triangular polygons. As you mentioned, I think it requires some iterative processing, and possibly scripting will be easier than using only existing transformers.

 

For example:

 

 

The TopologyBuilder will add "_polygon_id" to every polygon, and add "_left_polygon" and "_right_polygon" ("_polygon_id" of left/right hand side) to every line. The "_polygon_id" is 1-based sequential number (0 means "not existing"). 

 

This script example computes group ID ("_group_id") and creates features having a pair of "_group_id" and "_polygon_id".

 

Merge "_group_id" to the polygons and re-transform them into faces. You can then finally aggregate the triangular faces for each group.

 

-----

 

# Script Example for the PythonCaller

 

import fmeobjects

 

 

class GroupIdCalculator(object):

 

    def __init__(self):

 

        # List for set of polygon IDs for each group

 

        self.polygonIdSets = []

 

        

 

    def input(self, feature):

 

        left = int(feature.getAttribute('_left_polygon'))

 

        right = int(feature.getAttribute('_right_polygon'))

 

        if 0 < left and 0 < right:

 

            # Collect polygon IDs in the same group

 

            t , u = [], set([left, right])

 

            for s in self.polygonIdSets:

 

                if left in s or right in s:

 

                    u |= s # Unite sets

 

                else:

 

                    t.append(s)

 

            t.append(u)

 

            self.polygonIdSets = t # Update the list

 

            

 

    def close(self):

 

        for  groupId, s in enumerate(self.polygonIdSets):

 

            for polygonId in s:

 

                feature = fmeobjects.FMEFeature()

 

                feature.setAttribute('_group_id', groupId)

 

                feature.setAttribute('_polygon_id', polygonId)

 

                self.pyoutput(feature)

 

-----

 

Hope this works a treat for you.

There is always room for improvement in a script. 

 

This example may be a little more efficient.

 

-----

 

# Script Example for the PythonCaller: Version 2

 

import fmeobjects

 

 

class GroupIdCalculator(object):

 

    def __init__(self):

 

        # Key: Polygon ID, Value: Set of Polygon IDs in same group

 

        self.groups = {}

 

        

 

    def input(self, feature):

 

        left = int(feature.getAttribute('_left_polygon'))

 

        right = int(feature.getAttribute('_right_polygon'))

 

        if 0 < left and 0 < right:

 

            grp = self.groups.setdefault(left, set([left])) | self.groups.setdefault(right, set([right]))

 

            for id in grp:

 

                self.groups[id] = grp

 

            

 

    def close(self):

 

        groupId, finished = 0, set([])

 

        for id, grp in self.groups.items():

 

            if id not in finished:

 

                for polygonId in grp:

 

                    feature = fmeobjects.FMEFeature()

 

                    feature.setAttribute('_group_id', groupId)

 

                    feature.setAttribute('_polygon_id', polygonId)

 

                    self.pyoutput(feature)

 

                groupId += 1

 

                finished |= grp

 

-----

Sorry, forgive to mention that in the last scrrenshot, the orange selection is considered as a single feature.

GeometryCoercer has this parameter.

 

 

Would you like to play my with data?

I added three transformers - Chopper, Matcher, SpatialFilter. How's this?

 

Takashi,

 

 

Your solution seems to wok like a charm! I still however have some features not merged, but this is due to geometry (se screenshots 1 & 2). I think I can try to merge correctly using wall surfaces. Will try ...

 

 

Indeed, I am wondering why applying the same workbench using walls as inputs does not give a good result (see screenshot 3). Any idea?

 

 

Best regards,

 

 

Asher

 

 

 

Takashi,

 

 

One time more, thanks for your answer. From your screenshot, I can say that this is not exactly the result I am trying to obtain. The firsts results (where merged roofs are smaller: http://fmepedia.safe.com/servlet/rtaImage?eid=907a0000000csD3&feoid=Body&refid=0EMa0000000VCkw) is more the expected result. However, some roof parts are not merged, because they are not adjacent to other roof part, although they obviously belong to the same roof. I am however not sure this can be done.

 

 

This says, your results are very interesting sinc they improve to data. How did you managed to obtain them?

Removed the "WallSurface" and modified the process following the PythonCaller.

 

Other parts including the script are same as the previous version.

 

 

not elegant...

 

 

I think the resulting roofs are better than the previous results, but not perfect.

 

In this case, ideal automation might be difficult.

hello, maybe you will help me with my question..

 

I have shp file with attribute of height and I want to receive citygml file where walls and roofs are separated