Procedure for raster analysis

3962
3
07-23-2015 02:16 PM
Labels (1)
CaseyHofford
New Contributor

Hi, I'm trying to approximate distances between rock outcrops from aerial images and am not sure how to run an appropriate analysis on the raster data. Has anyone seen a similar paper with a procedure on this, or do you know what tools I should start w/? Thanks!

0 Kudos
3 Replies
DanPatterson_Retired
MVP Emeritus

Assuming you have the Spatial Analyst extension and a georeferenced image in projected coordinates, you can start with

Euclidean Distance—Help | ArcGIS for Desktop

DarrenWiens2
MVP Honored Contributor

Have you identified the rock outcroppings in your image?

If not, is it an RGB or multispectral image? If multispectral (or even RGB, potentially) start here for more information about image classification. I assume you'll want to go with supervised classification (i.e. draw some training polygons around known rock outcroppings, then extrapolate those spectral signatures to the rest of your image).

Once you've identified the outcroppings, run Region Group to identify distinct outcroppings. Then, convert the grouped outroppings raster to polygon. Finally, run the Near tool to get the distance from each outcropping to the next nearest outcropping (requires advanced license).

ChrisDonohue__GISP
MVP Alum

To tag onto what Darren mentioned, if part of your goal requires identifying the rock outcrops from the imagery, there are some things to look out for.

First, keep an eye out what imagery was used by others for this sort of analysis and what bands they used.  Often there are certain band combinations that have been found to work well for specific identification tasks.  Particularly important is finding out what interferences are out there that can cause misclassification.

Second, plan to do a few iterations before coming up with the final result.  I'll offer an example that is relevant given your topic - I used Landsat data from Eastern Washington State in the mid 1990's in grad school for a Remote Sensing class project to evaluate land cover.  Using Supervised Classification, I came up with a seemingly good classification system, but when I ground-truthed the results afterwards, it turned out water and basalt outcrops were both being classified as water.  And there were many basalt outcrops in my project areas....Ooops.  Ultimately, this led to using a different band combination to differentiate the two, with much better results in the long run, but it sure added to my workload in the short run to fix that unexpected outcome.

So keep an eye out for what combinations people use for outcrop identification and how they avoid interferences.  Additionally, when you do run the classification, be sure to include time for several iterations, with ground-truthing the results as part of each iteration if possible so you can refine your results.

Chris Donohue, GISP