GIS 4035 Module 5

This week we worked on Module 5 which had to do with supervised and unsupervised image classifications. In the first part of our lab we learned how to carry our unsupervised classifications of images in ERDAS Imagine. Along with practicing unsupervised classifications we also learned how to reclassify classes and how to merge classes as well. 

The second part of our lab this week introduced us to supervised classifications. We first learned how to manipulate spectral signature files and we later learned how to create our own spectral signature files of our own by using various methods. Once we created our own spectral signatures we learned how to quality check them by sing eucladian distance file images and also by creating and interpreting spectral signature histograms. 

As a final project for this week we were tasked with creating a land use map of Germantown Maryland by carrying out a supervised classification. We created our own spectral signatures, tested the signatures and used those signatures to carry out a supervised classification. Once the classification was complete we merged our classes and recoded them in order to simplify our resulting maps. The image file created by the classification Imagine was transferred into ArcPro and the map below is my final product. 

 

Photo Interpretation and Remote Sensing 

This week in class we learned about the different factors that could interfere with Electromagnetic Radiation (EMR) that is used for remote sensing. We also learned about and practiced some techniques to correct and enhance imagery. We added filters and manipulated bands in spectral imagery in both ERDAS Imagine and ArcPro. 

As part of our lab assignment we had to identify different features in a multispectral image that was provided. We used various techniques in ERDAS Imagine to identify certain features. Techniques we used to identify features included: visual examination of grayscale images, image histogram interpretation, visual examination of multispectral imagery (with varied band combinations) and the Inquire cursor. Identifying features ranged in difficulty as sometimes examining and interpreting an images histogram would give away a feature but in some occasions the spectral imager had to be closely examined to find the feature in question. Below is the map I created showing the features that I found for this lab assignment. 

Photo Interpretation and Remote Sensing Module 3

During this week's module we learned more about remote sensing. We got into a good amount of detail regarding the properties of Electromagnetic Radiation (EMR). We had a few math exercises where we determined energy levels, frequency and wavelengths of different EMR sources. 

For our lab assignment we started to explore and practice using ERDAS Imagine. ERDAS Imagine is software that helps extract different data from imagery. In our case we were exploring how to use ERDAS Imagine for remotely sensed data. We got to go over some basic operation such as loading data, manipulating data and preparing and exporting data from ERDAS Imagine to ArcPro. We extracted some data from provided imagery that contained land cover data that was classified in ERDAS Imagine. The extracted data was transferred to ArcPro where we created a simple map. The map I created is below. The map depicts land cover within a small portion of land within Western Washington State. 

 

 GIS 4035 

This week completed an exercise that calculated the accuracy of our photo interpretations. The first part of our assignment had us classify Land use and land cover using a true color aerial image. We created a new feature class and classified the entire area of the aerial image based on textures, patterns and associations. The second part of our assignment led us to test the accuracy of the land use/land cover designations that we made previously. We created a second feature class and digitized 30 points within the area of the provided aerial image. We used Google Maps to carry out ex-situ (offsite) ground truthing. The street views in Google Maps allowed us to see areas in a similar way that we would if we were actually in-situ (at the site). 

Once I completed the exercise I came up with only a 53% accuracy in my assessments. One of my major issues was that I classified a few areas as forest when they were actually just highly vegetated residential areas. My most accurate assessment was that of commercial and retail establishments. It was fairly easy to pick out commercial or retail sites. Commercial and retail sites are typically large buildings associated with large parking areas and large streets. Below is my resulting map.