Module 5: Unsupervised & Supervised Classification

In this week's lab, I classified satellite imagery using unsupervised and supervised classification methods in ERDAS Imagine. For the supervised classification, I created spectral signatures for 8 classes in Germantown, Maryland using the Region Growing Properties tool, Polygon tool, and the Signature Editor tool. After creating several Areas of Interests (AOIs) for each class, I examined the histograms and mean plots of all the signatures for spectral confusion. Based on the histograms and plots, I concluded that bands 4, 5, and 6 were the most separate and least confused of all the signatures. I used these bands for the supervised classification using Maximum Likelihood. I recoded the supervised image to merge the AOIs into 8 classes and calculated the area in acres for each of the classes. Below is the final result of supervised classification. 

Fig 1. Land use map from supervised classification of Germantown, Maryland.

Comments

Post a Comment

Popular posts from this blog

Module 2: Coordinate Systems

Image
In this week's lab I explore different coordinate systems and projections in order to better understand the nature of distortions introduced by the systems. For the last part of the lab, I was tasked with selecting a state in the contiguous US and selecting an appropriate projection for the area. I selected the state of Wyoming as my area of interest. Wyoming has 4 State Plane zones and two UTM zones, so neither a single State Plane nor UTM system would work for this area without distorting the size or shape of the state. Therefore, the best coordinate system for this area is Custom Coordinate system. I chose NAD 1983 WyLam (Meters), a custom system specifically for the state of Wyoming, as the projected coordinate system. This coordinate system is a Lambert secant-case normal-aspect conic type projection which preserves shape and is appropriate for a general reference map. The standard parallels are placed at 41N and 45N, the boundaries of the state of Wyoming, which minimizes
Read more

Module 6: Proportional Symbol and Bivariate Choropleth Mapping

Image
The objectives of this week's lab was to explore proportional symbol and bivariate choropleth mapping techniques, as well as customizing legends for each technique. One of the data sets for this lab included positive and negative values. The final map utilizes proportional symbols to show both positive and negative values. This map is essentially two proportional symbol maps overlaid on each other, one for positive job growth and another for negative job growth, with the symbols for both variables sized the same. Utilizing proportional symbols to show both positive and negative values is an effective way to visualize areas that show the greatest and lowest values. These areas are easy to see quickly on the map and give the map reader a simple way to visualize the variable being mapped. Fig 1. Proportional symbol map using positive and negative values.  Another data set for this lab examined contained two data variables: percent of the US population in each state that is obe
Read more

Module 4: Color Concepts & Choropleth Mapping

Image
In this week's lab, I explored using color and selecting color ramps in ColorBrewer, as well as applying these concepts to create meaningful choropleth maps. I created a map of population change from 2010 to 2014 in Georgia counties. I normalized the data and produced a choropleth map of population change using a diverging color scheme with 6 classes divided around a midpoint of 0. Fig 1. Choropleth map of population change in Georgia counties using a diverging color scheme. For the Georgia population data set, I chose to divide the data set into 6 classes around a midpoint of 0: 3 showing a decrease in population and 3 showing an increase in the population. The classes were grouped based on a manual classification closely based on the Natural Break classification, where the midpoint was adjusted to 0 with the other groups unchanged. Natural Breaks classification creates classes based on similarities in values in the data, thereby revealing patterns in the data. By adjust
Read more