Introduction
In this lab I used an online tutorial to calculate engage in "Value Added Data Analysis." I essentially used ArcPro to calculate the aerial impervious surface area of a given map. The tutorial that I used can be found below.
https://learn.arcgis.com/en/projects/calculate-impervious-surfaces-from-spectral-imagery/
Methods:
This lab is a bit odd in comparison to the others in that it relies on sampling and AI to perform calculations. In Figure 1 you can see the starting image, apparently a neighborhood near Louisville, Kentucky.
 |
| Figure 1: Pretty standard map |
Segmentation
I started off by segmenting the image for easier classification. To do this, I had to start by extracting the spectral bands by using the "Raster Functions" in the "Imagery" tab seen in Figure 2.
 |
| Figure 2: Raster Functions |
Using the extract bands function found in raster functions and the combination "4 1 3" the original image was converted to what can be seen in Figure 3.
 |
| Figure 3: Extracted Bands |
After that, I had to set up the image classification wizard in preparation for the segmentation. Skipping over the specifics of the classification wizard (which can be found in the linked tutorial), the spectral detail was set to 8, spatial detail to 2, minimum segment size in pixels was set to 20.
Spectral detail is the difference that adjacent pixels have to be from each other in color in order to be grouped together. A higher number, on a 1-20 scale, will create more, smaller, segments. A smaller number will create fewer, larger, segments.
Spatial Detail is very much like spectral detail except that it is based on distance between similar pixels instead of color. It is also on a 1-20 scale with higher meaning more segments.
Minimum segment size in pixels is the size that each segment must be before it is instead merged into a neighboring segment.
The resulting segmented image is shown in Figure 4.
 |
| Figure 4: Segmented Image |
Classification
The next step is classification of segments. By manually selecting colored segments you can classify what exactly that segments is, whether it's road, grass, water, or rooftop. Figure 5 is an example of my classification.
 |
| Figure 5: Classification |
This classification process was done in the classification wizard training samples manager page. I created two classes, pervious and impervious, and broke them down further into the things that can be found on the image. The magical wizard AI is then able to discern what is pervious or impervious based on my specifications. The image was first, however, converted to a intermediate stage where the samples were separated (figure 6).
 |
| Figure 6: This image is classified |
From there the image can be converted to a full separation of pervious and impervious as seen in figure 7.
 |
| Figure 7: Grouped Classification |
