# Create a raster that measures distance of two polygons

I'm currently working on a project where I have 2 datasets/layers that comprises boundary polygons (vector) of farms. One has been digitized from community mapping (overlaying a UAV image) the other one is collected using ArcGIS Collector. Now what I want to do is to know the distance between these two layers. For example Farm A boundary using UAV vs Farm A boundary collected by ArcGIS.

My goal here is to provide summary statistics on how accurate and fast is it to delineate polygons using a geo-referenced image without having to go to the site itself and collect data.

I'm using ArcMap 10.5 and wondering what tool (ArcGIS toolbar) I can use for this? My co-worker suggested that I can create a raster that would measure the distance between the two. Then clip the raster and calculate summary statistics.

I'm not really sure what to do based from her suggestion

As an alternative I would calculate the the union of the two features.

I would then calculate the intersection of the two features.

I would then clip the union by the intersection.

The remaining piece is the area of difference. You can do this method with Vector layers.

• I have done this alternative so far and it worked but I'm still hoping to do something about the raster option. I'm trying to work out if it's still needed and trying to find a way wherein we won't need any code to do this. But thank you as this has been very helpful! – Janeen Kim Cayetano May 22 '17 at 8:43

It seems to me that you will need two calculations. First you would create binary rasters. Use a value of 1 for the area covered and 0 for the area not covered.

Using the full extent of your features, clip the rasters to minimize the calculation.

Finally sum the absolute difference between rasters. I think you will need to create a script to do this. Use the absolute value because it will give a true indication of error when the field value is false and the automated value is true and vice versa. If you just sum the differences you would be offsetting one bias vs the other.