GIS 5027 Module 5 - Unsupervised and Supervised Classification Methods

 

This week's lab assignment focused on Unsupervised vs. Supervised Classification methods. The first part of the lab explained how to perform an Unsupervised Land Use / Land Cover Classification on an aerial image of the University of West Florida campus. The second part of the lab walk us through the Supervised Classification method. This method is much more extensive and methodic, creating a more accurate classification map. As displayed above, our final deliverable for this project was a land cover / land use map of Germantown, Maryland. While I am happy with the quality of this map, I would like to know for sure what accuracy level the classifications achieved; not knowing this information concerns me to an extent. Also, if I had to redo this project, I would definitely change the initial band combinations of the original satellite image. For the map above, I went with a false color Red(4), Green(3), Blue(2) band combination. Other than these minor issues, this laboratory assignment was extremely informative and a nice set-up to dive into the final project...

GIS 5027 Module 4 - Spatial Enhancement, Multispectral Data, and Band Indices

 Module 4's Lab Assignment was focused on downloading satellite imagery [through the United States Geological Survey] and making adjustments to these multispectral images to make areas of concern more visible to the human eye. This was first accomplished through the use of low-pass filters, high-pass filters, and sharpening filters. After exploring these options in ERDAS Imagine, we applied various filters to images in ArcGIS Pro. This exercise was brief in nature, so exploring the differences between the filters included in both programs will be necessary to differentiate between the strengths, weaknesses, and outcomes of each. The next portion of the lab was targeted at investigating brightness levels on the different bands of the satellite imagery by investigating each layers histogram; this can be done in both programs [Imagine and ArcGIS], but I chose to focus on Imagine and import the subset images into ArcGIS to format the map layout. For this exercise, we were tasked with identifying three features based on spikes in pixel values on specific layers of the multispectral image.

The first feature we were asked to identify was one that had a spike in pixels with very low brightness values; given the fact that water absorbs a lot of electromagnetic radiation, this one was relatively easy to discover. As seen in the map above, I chose a band combination that gave the water a very dark [almost black] tone that greatly contrasted from the bright green / orangish tones of the land surrounding it. 

The second feature we were asked to identify reflected almost all of the visible light, but absorbed electromagnetic energy in the thermal spectrum. This indicated the presence of ice / snow. In the map above, I chose a band combination that gave the snow / ice a bright pink tone and the land surrounding it a contrasting brownish orange tone. The green areas would be parts of the ice / snow that are melting, or areas that are reflecting energy that is closer to the near-infrared spectrum.

Finally, we were asked to find a band combination that would exaggerate the areas of the water that were reflecting higher levels of visible light [possibly caused by the presence of sediment in the water that would be reflecting the electromagnetic radiation, or possibly areas of water that are more shallow than the darker areas]. As displayed above, I chose a band combination that gave the water and land a moderate contrast, but the yellow areas within the water provide the high contrast needed to investigate the areas in question. 

Overall, this lab was very intensive and full of information, but was a good opportunity to really begin investigating the spectral signatures of various features / materials and also a great opportunity to learn the capabilities of ArcGIS and ERDAS Imagine as well.

GIS 5027 Module 3 - Introduction to ERDAS Imagine

 Module 3's Lab Assignment was to familiarize ourselves with a new software program named ERDAS Imagine. While there are a daunting number of options and buttons, I feel like this was a proper introduction to a complex program. The map above is a subset of a larger raster image of a portion of Olympic State Park [located in western Washington state] that displays each pixel as one of seven different categories defined by the reflectance obtained by AHVRR satellite imagery. The objective of this assignment was to compare the areas of the entire raster image with the areas of the subset image that was created from the original. As displayed in the legend, there are [7] different classifications and the area of each [in hectares] is displayed next to the category name. Comparatively speaking, the areas of the classifications listed in this subset are minute compared to the areas listed in the original raster image. Finally, the subset image was imported into ArcGIS and a map was created to aesthetically display the information obtained throughout this assignment.