Final Report for GIS 4035 - Photo Interpretation & Remote Sensing

I have definitely lost my mind...but I know where it is. My final project for GIS 4035 - Photo Interpretation & Remote Sensing, was to apply remote sensing data and image processing techniques that I studied and used during this semester. I utilized concepts, data types, processes, and techniques I learned in ERDAS Imagine, and ArcMap. What an experience this has been! My project was to examine the Lake Tahoe basin area and determine the change, if any, in healthy vegetation from the period of 1999 to 2010. The guiding question for my research is: How much has healthy vegetation decreased in the Lake Tahoe basin area from 1999 to 2010?

It was a fun project and I certainly have learned a great deal this semester. Please see my Final Report at the link below. Here is my final map:

Lake Tahoe:1999 to 2010 -Decrease in Healthy Vegetation

Supervised Classification: Germantown, MD

What a week it has been. In Photo Interpretation, we continued learning about classification of aerial images. This week was Supervised Classification. Using ERDAS Imagine, and a raster image of Germantown, MD, I created my own Spectral Signature and Supervised Classification. I used the AOI Seed tool (Drawing tab, Insert Geometry group, press the Grow dropdown arrow, then click Growing Properties) to designate eight areas (pixels) on the image that were of a particular type of land cover, including:

Urban/residential  

Grasses 

Deciduous Forest

Mixed Forest (Deciduous/Conifer)  

Fallow Field

Agriculture

Roads

Water

I used Mean Plots and Histograms to evaluate where there was Spectral Confusion. More accurately, I was looking for areas where the Mean Plot lines did not converge or which bands had the most separation. I discovered that my Plots had the most spectral separation in Bands 1 -2- and 3. Then I used the  Maximum Likelihood tool to classify my image after setting color to "Approximate true color."   

The result of my work including a Distance Map is shown below.

Unsupervised Classification

This week I learned about the two ways to classify pixels on an image: Supervised and Unsupervised. For a Supervised classification, I would identify "training pixels" to use to identify the rest of the features on the image I was working with. For Unsupervised classification, the pixels are simply divided into a number of classes depending on the number you specify. The tools to accomplish this in ArcMap are: Iso Cluster Tool and the Maximum Likelihood Classification tool

We also used ERDAS Imagine to classify pixels. In ERDAS, the classification process was more tedious but generally worked the same. In the Raster tab, under the Classification group, click the Unsupervised button, then select Unsupervised Classification.

I performed an Unsupervised Classification using an aerial of the UWF Campus. I made five classes, named them and designated colors for each class:

1. Grass – green  

2. Trees – dark green 

3. Buildings/Road – grey 

4. Shadows – black 

5. Mixed - orange

The process was long but I am happy with the results. Additionally, I created another column or field for "Area" and then I totaled the area as "Permeable" or "Impermeable" and listed the percentages of each on my final map.  

Thermal & Multispectral Analysis - Barren Soil, Guayaquil, Ecuador

For this lab I conducted an analysis of a particular feature using Thermal & Multispectral tools. The primary goal was to use both ERDAS Imagine and ArcMap to adjust band combinations to highlight a feature of my choice. I created a composite image in ArcMap using a LANDSAT ETM image of Coastal Ecuador.  (Data Management Tools > Raster > Raster Processing > Composite Bands). At first I examined each band separately. For Band 6, the Thermal Band, the image remained mostly the same as in all the previous bands, only it appeared more “fuzzy.” The thermal return indicated that my feature was quite warm. This told me that it was a material (bare earth) that was emitting a good deal of heat. Knowing that I was viewing a feature very near the Equator, I surmised that this feature was emitting retained heat from the sun. I could also venture to say that this image was not taken at first daylight since the bare soil was already at least "warm".  I would guess the bare soil had absorbed at least some sun energy and therefore the time of day was at least past mid-morning. Additionally, I could not detect any significant shadows so this also told me the sun was probably at a high angle overhead. After viewing each band separately, I used various band combinations to further examine the image including:

False color – 4-3-2

Natural color – 3-2-1

Healthy vegetation – 4-5-1

Near IR, Mid IR, Red – 4-5-3

I finally decided to use the Band combination of 7-4-2 to display the final map analysis. This combination provides a "natural-like" appearance of many features. Healthy vegetation is bright green, grasslands appear green, pink areas represent barren soil, oranges and browns represent sparsely vegetated areas. Hence, the feature I selected in this map of Guayaquil, Ecuador is a patch of barren soil surrounded by healthy vegetation and water.

Multispectral Anaysis

This week in Photo Interpretation we continued our study of image enhancement with multispectral analysis. This was a great week where I learned how to use:

1) Examine the histogram for shapes and patterns in the data.
2) Visually examine the image as grayscale for light or dark shapes and patterns.
3) Visually examine the image as multispectral, changing the band combinations to make certain features stand out. (This one was a blast!)
4) Use the Inquire Cursor to find the exact brightness value of a particular area (The Inquire tool was my favorite).

I worked through several tasks in ERDAS Imagine to accomplish the above.  The first Exercise was using Histograms in ERDAS. Manipulating Breakpoints was a challenge for me. Though I understand the principle, getting the correct image was difficult. Exercise 2, was Using Spectral Characteristics. My main take-away is:

a.      The band combination, Red: 4, Green: 3, Blue: 2 is called

      Near Infrared.

b.      The band combination Red: 5, Green: 4, Blue: 3 is called

      Short Wave Infrared.

c.       The band combination Red: 3, Green: 2, Blue: 1 is called

      True Color.

These band combinations help to highlight certain features and make it easier for us to find or identify important or otherwise, elusive features on a map. Exercise 3, Band Ratios - Creating Indices was very interesting to me. I created a Normalized Differential Vegetation Index (NDVI) that helps to distinguish clearcut areas. I could have used this tool for identification of MTR Mining sites in Special Projects!

The last Exercise, was to find 3 "Mystery" features matching pixel criteria as specified below.

The first feature was:  In Layer_4 there is a spike between pixel values of 12 and 18. Name the type of feature responsible for this and locate an example of it on the map. I used all four of the techniques above and with some trial and error I identified a large body of water as this feature:

I selected a True Color band combination to display this feature.

 The next "Mystery" feature was: A small spike in layers 1-4 around pixel value 200, and B) a large spike between pixel values 9 and 11 in Layer_5 and Layer_6. I selected a snow-covered mountain range as this "Mystery" feature.

I selected a Short Wave Infrared band combination to highlight the snow on these mountains.

The "Mystery" feature here is Mount Olympus in Washington State. The mix of dark, small, bare rocks on the mountain tops made this identification more challenging.

The third and final "Mystery" feature was to identify a certain type of water feature: Layers 1-3 become much brighter than normal; layer 4 becomes somewhat brighter, and layers 5-6  remain unchanged. 

Using the four techniques from above, finding this water feature was a difficult hunt. I thought the type of water might be near shore, so it would be a mix of rocks, shallow water and perhaps some turbidity. But the more I searched using the Inquire tool I became convinced the feature could be a stream or river. The Queets River, along its banks seemed to exhibit the above traits. So, this is the feature I selected.

  

I displayed this feature using a Near Infrared band combination, 4-3-2, The Queets River in Washington State, is located on the Olympic Peninsula, mostly within the Olympic National Park. This river empties into the Pacific Ocean. 

Lab 6: Spatial Enhancement

This week in Photo Interpretation we took a look at downloading data from USGS to obtain a Landsat Archive of Landsat 4-5 TM. The objective of this lab was to use various image enhancement tools and techniques to improve the quality of your image.

I used ERDAS Imagine to begin my spatial enhancement. The first technique I used was the Fourier Transformation. A Fourier Transformation is a mathematical technique for separating an image into its various spatial frequency components. I had a Landsat image that contained striping and the Fourier Transformation afforded me a method to remove that striping. The next tool I used was the Wedge button. This was a challenge to get the correct amount of coverage and not delete too much detail of the image. I also used the LowPass button that seemed to smooth the image a bit more. After completing these tasks in ERDAS Imagine, I switched to ArcMap to try a few more tools to enhance my image. I tried other filters and changing the brightness and contrast. In the end, I built pyramids for my images and this seemed to add a slight bit more contrast to the image. Below is my final map.

ERDAS Imagine - Remote Sensing & EMR

This week I learned about ERDAS Imagine.  This is a remote sensing application for GIS. The ERDAS Imagine program allows us to manipulate raster images to extract information. In this lab we were using a raster of Washington State to classify the types of land cover such as vegetation, water, bare earth, etc. To do this, initially we began with some calculations of electromagnetic radiation (EMR). The chart below shows the electromagnetic spectrum:

There were a few formulas involved such as Maxwell’s wave theory: C = λν and Planck Relation: Q = hν 

But the real point of this lab was to learn about ERDAS Imagine and how to process a raster image to understand the types of land cover depicted. A very useful feature was the ability to switch between TM False Natural Color and TM False Color IR. The False Natural Color gives the bare ground a pink color which adds a sharp contrast making it easier to discern from the green forested areas.

Overall this was a fun Lab but I am sure I will need more practice with ERDAS Imagine to feel as comfortable with it as I do with ArcMap. Below is the map I created for this week's Lab. It shows an area in Washington State that I have re-classified by using the Inquiry Box and manipulating of the ERDAS Attribute Table.

  

Ground Truth: Error Matrix & Overall Accuracy

This week I modified my map from last week, Land Use Land Classification. The challenge last week was to not use any tool that would help you verify the type of land I was classifying. I could only use the aerial image (.tif) of Pascagoula, MS. and make my best guess at what type of land I was classifying. This week, the challenge was to take an unbiased sampling of the land I classified and compute the overall accuracy. Under normal circumstances, you might travel to the location and conduct fieldwork to determine the correct land use/classification by conducting a Random, Stratified Random, or Systematic Sampling. Since I could not actually travel to Pascagoula, the tool of choice was Google Maps - Street View. This is an excellent proxy for actually being there.

Another tool I created and used was an Error Matrix. I conducted a virtual Random Sampling of my map starting at the North West border and working my way South to cover the entire study area and I marked as "True" or "False" the areas I had previously classified. I used Google Maps to confirm what I classified was correct - or not. I found that I miss-identified a few land types such as:

- A Commercial / Industrial area (Category 15, USGS Andersonian framework for national land use and land cover) - What looked like an Industrial area to me turned out to be a shopping center.
- A Residential area (category 11, USGS Andersonian framwork) - I correctly identified most of the residential areas, but one area I classified as residential turned out to be a cemetery!

My Overall Accuracy was 76%. Here's the map I created.

   

Land Use Land Cover: Pascagoula, Mississippi

Another fun week! This week I worked with the USGS Andersonian system for classification of  Land Use Land Cover. I continue to hone my skills for using, Shape/Size, Shadows, Patterns, and Association, for identifying various land features. In this exercise, we used the USGS Anderson framework for naming Land features. The area we examined was Pascagoula, MS. This location has loads of water features. Digitizing the water or the river, was my biggest challenge. I took a large cut at trying to draw my initial polygon around as much water as I could only to have the polygon distort or completely disappear. I finally got the Create Features to work by digitizing smaller areas at a time. This work can be tedious, but I enjoyed it very much. As it happens, I am working on a similar project at my internship with Brevard County, where I am digitizing and updating storm drain pipes throughout the county.    

GIS 4035 - Visual Interpretation

This has been another busy week. The Lab for Photo Interpretation and Remote Sensors was actually a three for one.

The first Exercise for Visual Interpretation was to use Tone and Texture to discern various features using an aerial photo. I used an aerial photo that showed features such as a runway, housing subdivisions, water and vegetation. Tone is defined as the  brightness or darkness of an area and the first part of the lab had us classify areas from very light, to very dark. Using the "Convert Graphics to Features" went smoothly for the first Exercise, but I had some technical challenges for the second Exercise. Another objective of this lab was to look critically at the features you are given in an aerial photo so that you can distinguish the subtle differences between organic and man-made features. Below is the map I created for Exercise 1:

  Exercise 2 consisted of Identifying Features by using Shape & Size, Shadows, Patterns and Association. The objective here was to use all available information to ensure you correctly identified the object or feature you were looking at. As I said above, I had some technical challenges with part 2 with the "Convert Graphics to Features" option. It  seemed as though each time I needed to convert my selection (Shape & Size, Shadows, Patterns and Association) the option was greyed-out and not available. I tried Starting an Editing session and Stopping Editing; I tried opening and closing AcrMap. I walked away for a while and when I came back, the Convert option was still not available. Somehow, I did get the Convert Graphics to Features to work....mostly. But I do not know why the option was greyed-out.  Below is my map for Exercise 2:

The last portion of the Lab, Interpreting Color, had us use two .tif's to compare the indicated color of features using True Color and False Color. False Color is also called Color Infrared. I found various features including land, water, healthy vegetation and distressed vegetation. The colors I found matched closely, but were not exactly the same as what the "legend" for "False Color IR vs True Color" stated in the Lab instructions.