Removing duplicate features with same geometry in ArcMap?

Question

Is it possible to uniquely identify each geometry in a feature layer?

Essentially I have a polygon feature layer with many rows of data represented by the same geometry. I would like 1 row for each unique geometry. So I'm looking for some out-of-box methodology to combine attributes based on the shape of the geometry.

I come from a Postgres world and can solve this problem there, but I don't know how to do it (or if it's possible) to do it non-programmatically with ESRI tools.

Answer 1

If you have an advanced license, you can use Find Identical or Delete Identical.

Both can be used to find/delete features that have identical attributes, or, if the Shape field is specified, identical geometries.

If you don't have an advanced license, Removing spatially duplicate features using ModelBuilder? will be useful.

In short, you add two fields for X,Y in your attribute table and run a Dissolve on said fields.

If you have polygons which share the same centroid, but are not geometrically identical, you can add two other fields (Perimeter and Area). That should be enough to identify identical geometries for almost all normal cases. See @whuber's comments below.

Answer 2

I have a dataset based on survey data. My problem is that old features do not always get removed before the new survey shots are imported. Therefore we have 'duplicates' with different geometry: IE the survey shots might be 1/4 ft separate from each other. On a 12000 feature dataset this is difficult to zoom into each and identify them. Using ArcMap Basic 10.2: here's a hacky model I came up with.

1. Buffer(w/input radius) each feature in original layer (Selected or Not)
2. Iterate over each feature in buffer layer
3. Select from original layer based on relationship (Centroid w/in feature)
4. "Row count" selected features
5. Field Calculate Row count value to buffer layer = How many features are within radius of original features

I re-invented the wheel. Also, I got frustrated with Model builder and basically bulldogged the whole thing. There has to be a more elegant solution. I will gladly accept critiques and suggestions.

Also, this seems to only run correctly from within Model Builder, I run from a toolbox and it does not save the buffer layer.

Answer 3

I have been looking for an answer to the same question for some time to overcome flattening the overlapping buffer polygons issue and think found a robust solution to that. In fact @Whuber's comment,

In principle the centroid, perimeter, and area might still not uniquely identify a polygon, but the counterexamples are a bit contrived.

on @Paul's answer was the main drive for me to keep trying. It uses Spatial Join tool's ARE_IDENTICAL_TO operator since this is the most efficient among other identical detectors. Here are the steps:

• Self Union your input to keep all shapes different than each other;
• Create a dummy field to keep/copy original OBJECTID's (i.e., OID field type) of the rows, say EX_OID;
  • Run Spatial Join on the feature class that is wanted to be processed as both target and join feature parameters. Make sure the field mappings of dummy field(s) (appears twice) previously created has the merge rule of FIRST (which is the default or pick LAST, the others may not give the desired output since they use a summary of overlaps). This step picks the common shape OBJECTID sitting on top of all overlapping (i.e., FIRST);

Following steps can be done in two ways, WITH Python or WITHOUT:

WITH PYTHON

• Use the code below to extract unique IDs that we want to find:

print tuple(set([row.getValue("EX_OID_1") for row in arcpy.SearchCursor("Output_of_the_Spatial_Join_Operation")])).

Be aware that I am using the second copy of dummy OID field, which is EX_OID_1 in my case;

• Copy the output of this code and paste this into the Definition Query of the original feature class processed as OBJECTID IN (<PASTE_COPIED_VALUES_HERE>). The result will be showing you the unified features of the overlapping parts. There is a size limitation here though, 30,000 characters. To overcome this issue, you need to follow "without python" steps.

WITHOUT PYTHON

• Use Frequency tool to list unique values in the second copy of dummy OID field, which is EX_OID_1 in my case;

• Lastly Join OBJECTID field of the original feature class with the output of the Frequency tool's EX_OID_1 by selecting "Keep only matching records" options.