I have a vector layer (Originally a raster layer) of well sites, cut blocks, and seismic lines all lumped together in the same feature class and no distinguishing attribute between them. enter image description here

I want to split these linear polygon features from the non-linear polygon features. For Example, I would like the polygons to get split where I drew in the yellow lines.

enter image description here

Is there an approach to do this in either vector or raster data format? I don't care about maintaining field data.

I am using ArcGIS 10.2 but am open to solutions with any program.

  • This is very tough, I had a similar file once. You may want to exam non-GIS image software such as Photoshop & Gimp. in Arc you may be able to "nibble" at these or maybe even some form of Kernel but it will be far from clean. I hope someone has a good solution. Sep 22, 2014 at 22:39
  • Thanks. I actually want to isolate them because I need to nibble the linear features but not the non-linear features and it's to large a data set to do it manually.
    – Calavin
    Sep 23, 2014 at 15:01

1 Answer 1


I think you should process your raster using the Generalization toolset of the Spatial Analyst toolbox.

A Shrink by a value just large enough to remove those linear features, followed by an Expand of the same amount prior to your raster to vector conversion should do the trick to isolate the polygons.

The difference between your polygon raster and the original raster will be the linear features.

  • Accurate vector approach: convert polygons to polylines, append with split lines. Planarise and build polygons. Raster: convert split lines to raster, assign no data values to original raster, where split raster is not null. Convert to polygons.
    – FelixIP
    Sep 22, 2014 at 23:00
  • Thanks for the comments. I have tried the shrink and expand approach with a cell value of 2. It seems to have got rid of all the linear features. Not all of the non-linear features expanded back to its original extent, but I think that is something I can live with. I then reclassed the new expanded raster so that the features that remain = 1 and NoData = 0. I did the same with the original raster and used raster math to add both rasters together. Now I have a raster with values of 0 = NoData, 1 = Linear features, and 2 = Non-linear features.
    – Calavin
    Sep 23, 2014 at 19:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.