AnsweredAssumed Answered

Combining shapefiles for land classification by consensus?

Question asked by ShackletonTea on May 6, 2020
Latest reply on May 31, 2020 by ShackletonTea

Manual classification of satellite imagery can be difficult due to the complex and often non-distinct boundaries between different types of landscapes. As a result, the manual delineation of boundaries can vary between annotators 

 

Therefore, I have asked a number of volunteers to map a single satellite image of the coast using polygons. I wish to combine all of their shapefiles to determine the final polygons for each class.

 

There are three classes (sand.shp, water.shp and vegetation.shp).  

 

Due to human error, individual annotators will often have "gaps" where the boundaries of two polygons have not snapped or "overlaps" where two different polygons (e.g., "sand" and "water") will overlap. 

 

What I would like to do is combine all of their shapefiles into a final map, whereby if the majority of volunteers shapefile's agree for a given region, that region will be classed. If there is a disagreement between volunteers, then the region is labelled as "non-classified". 


What is the best process of doing this? 

 

I imagine I could convert the shapefiles to raster and do a raster calculation to class each pixel by a threshold, where if >70% of users agree on a class it is that class. If not, then class as "non-class". However, I would like to know if it's possible to do this by another, and perhaps more simpler, means by directly working with shapefiles and utilising ArcGIS Pro /ArcMap tools? 

Outcomes