Smart Classification using ArcGIS Pro
In my last post, I discussed how we can use the color green as a proxy for plants for vegetation analysis. Sometimes simply classifying plants vs. developed or barren surfaces is enough. Other times however, you might want to differentiate between herbaceous plants and conifers. If your map is small enough, you can hand draw polygons around each section, but for a large map or an area with lots of convergence, smart software can help tremendously. Take the meadow below for example:
This meadow has a river running through it with gravel banks, bounded by a road and trees along the periphery. For this project I used ArcGIS Pro’s Image Classification Wizard which has two modes: supervised and unsupervised. In the supervised mode, you draw polygons for each of the features and the software attempts to match those sets of pixels to similar sets of pixels in the map. In the unsupervised mode, the software creates a set of classes based on color (and maybe shape?). In this example I had it create 15 classes, which I then reclassified using the NLCD (National Land Cover Database) categories. The reclassification is quite simple in ArcGIS Pro, you can quickly toggle between the original image and the new classes, selecting all the classes which are water, shrubland, wetlands, etc..
Once your images are reclassified, you can do a dissolve on the new category to do some statistical analysis, for example what is the percentage that is wetland vs. herbaceous. I don’t use ArcGIS Pro often so I’ve only just begun using this tool. I’ll update this post or write new ones as I learn more about this powerful tool. One challenge I ran into with this first meadow map was that shadows of taller trees became a class in the set, but shadows might cover very different land types. You might remember that I flew this meadow early in the day (9am), so a midday flight might help in this regard (shorter shadows). I’ll try using the supervised classification method next and specifically see if that helps with the shadowed areas.
