Week 5: Supervised Classification

Hello all!
This will be a short explanation because I have been burning the candle from both ends for several days and I'll soon crash.
So this week I preformed a supervised classification of land use for Germantown, Maryland using Erdas and its supervised tool.
An AOI image was added to the provided aerial image, and using a mix of the Inquire (legacy) cursor, region growing properties tool, signature editor, offered coordinates and classifications, a heavy dose of eyeballing and referring to examples classified the land use. Then I recoded the classes to be a single number, re-added the names of the classes that correspond to the number, and add area.
It sounds easy but between being so tired and being slightly unsure if I was doing things right I re-did and deleted things several times until I was satisfied. Which only took 7 hours just for the last exercise before I transferred the image to Arc GIS Pro and created the map which is below.

Week 4: Image Preprocessing

Hello all,
 This week ERDAS Imagine was used to process and look at a raster image and identify 3 different aspects of said image.
Feature 1: Using the inquire tool I found that a spike in pixels between 12 and 18 in Layer_4 is water. In my opinion it was easier to see the Water in the raster band combination Red: 4, Green: 3, Blue: 2, or near infrared because water shows up as a dark blue or green in this band combination whereas the land is a range of red colors.
Feature 2:  Using the inquire tool and histogram I found that Snow on the mountain had small spike in layers 1-4 around pixel value 200, and a large spike between pixel values 9 and 11 in Layer_5 and Layer_6. I used the  raster band combination 1, 1 , 1 or grey scale to make the brightness in the pixel values to shine through the most.
Feature 3:Using the inquire tool and histogram I found that the water of a small lake has much higher pixel values in layers 1-3 while layers 5 and 6 remained relatively unchanged. I used the true color combination or Red: 3, Green: 2, Blue: 1 to make the bright blue reflection of the lake to be seen the easiest.

After these features were found I created a subset of the image then imported the images to create a map for each of the features in Arc GIS Pro which are below.

Week 3: Intro to ERDAS

Greetings on this chilly blustery day,
So this week was an introduction to ERDAS Imaging 32 bit. It was not all that bad for a first time, the interface reminded me of windows 64 a little bit but it functioned pretty smoothly. I can't say that I like the little round about ways to save or work on things but I like the save session option. I can see its applications and favor ability but it will take a while to get used to the new program.
The task this week was to navigate around ERDAS and get comfortable with the program. Then take a provided raster image, edit the bands to be a TM false natural color band combination, add an area column to the attribute table then crop the image to size using the inquire option, that the image and export it to Arc GIS Pro and make a map with legends for area size and colors.
This is said map.

Week 2: Land Use/Land Cover

Hello again,
       This week was about land use and land cover where I took a provided image and made a feature class (LULC) to create polygons in order to class II classify the land usage of the image using ArcGIS Pro. The levels were as follows: 

1. Urban or built up land
1. 11: Residential
2. 12: commercial service
• 129: vacant
3. 13: Industrial
4. 14: Transportation, Communications, and Utilities
2. Agricultural Land
1. 21 Cropland and Pasture
3. (4)Forest Land
1. 41 Deciduous Forest Land
4. (5)Water
1. 51 Streams and Canals
2. 52 Lakes
3. 54 Bays and Estuaries (did not have enough time)
5. (6)Barren Land
1. 73 Sandy Areas other than Beaches

After the digitizing and categorization was done I created a point shape file to plot points at semi-random to check for accuracy. My method was kind of used a grid where I went left to right and clicked at random on that line motion. Then to check I used an "in/on field" accuracy assessment using google maps street view. I was unsure how to calculate accuracy and I forgot to add the accuracy to the map. But I was unsure of the numbers anyway. If we count the water areas I did not digitize then I probably got around 60-65%, and if we ignore that I think I would say 70-75%. I may be wrong however. 

     The colors I chose to categorize was from the random assortment of bold colors with slight tweaking so green corresponded to trees and blue to wetlands and all set at a transparency of 40-50% depending on the vibrancy.

     Below is the finalized map, I probably spent too long on the digitizing but it looks good.

     Thank you for reading and hopefully you stay tuned for next week.

     ~Jo Snow.