Module 4: Image Preprocessing

In this module, I worked on spatial enhancement, multispectral data, and band indices.

Exercise 1:

    This part was acquiring the satellite data for this lab. I navigated to glovis.usgs.gov and found the Landsat 4-5 TM image of the Pensacola Bay area.  The image data I got was a little different than the example in the lab instructions because the database is continually updated. At the beginning of this lab, I had a bit of a problem extracting the LT50200392011045EDC00.tar.gz file. I found out this was because the computer I was using did not have 7-zip. Once I switched to “GIS Files” I was able to extract the files appropriately.

Exercise 2:

    The different spatial enhancements discussed in this lab are interesting. Kernel high pass filters suppress low-frequency data and enhance data that changes from pixel to pixel (edges, lines, buildings, roads). Kernel low pass filter suppresses data that changes from pixel to pixel and enhances data that does not change much between pixels. Fourier transforms were introduced but not performed in this lab. 

    Using basic high/low pass filters in ERDAS Imagine was pretty straightforward. Under the Raster tab, I used the “Spatial, Convolution” tool to recalculate the image. It was good to see the differences between the original image, low filter, and high filter. The sharpen filter looked very similar to the high pass filter but to me, the river edge looked better in the sharpen filter image. 

    Task 2 was to use other filters in ArcPro. I created a Mean7x7 filter following the lab instructions. This created an image similar to the low filter in ERDAS Imagine. This one is more generalized because it uses a 7x7 kernel, not a 3x3. The Range3x3 filter was very interesting because it made it very easy to see borders between different features. 

Exercise 3: 

    This section of the module covered image histograms. Reference the lab instructions for information on how to read the ERDAS Imagine Metadata-Histogram graph. It includes information on the brightness value, frequency, and minimums/maximums. The first task was to adjust the breakpoints and LUT histogram. This task was confusing to me and I had to read the directions several times to understand what was being asked. I moved the breakpoints so the LUT histogram stretched across the entire DN Histogram. This excluded some of the higher DN numbers and made the image lighter. I then performed a similar histogram function in Arcpro. 

Exercise 4:

    In this step, I adjusted different spectral characteristics to help identify features. I opened tm_00.img again and changed it to Landsat 5 TM-6 bands with a TM False Color IR band combination. I navigated to the coordinates in the lab instructions using the inquire function. This location was a clear-cut area. Switching the band combination to TM False Natural color did not significantly improve the visibility of the logging roads. In my opinion, the clear-cut areas did stand out better in this combination but not by much. The True Color band combination appeared similar to how we would see it. Adding Band 1 in a second view somewhat improved identification of clear-cuts/paths but some of the detail was lost. Bands 1- 3 appeared most helpful in identifying clear-cuts and trails.

Exercise 5:

    This section covered creating a Normalized Differential Vegetation Index (NDVI). This was done using the ERDAS NDVI tool. This made the clear-cut areas very visible. They had a near-zero value and were very dark in the image. 

Exercise 6:

    I examined the layer metadata in this exercise. I used the histogram to examine spikes in different layers. I had to reference information earlier in the lab to be able to read the graph. Information on the pixel value can also be obtained using the Inquire Cursor (which we have used before). I am still confused about what the LUT Value is even after researching it on my own.

Exercise 7:

    In matching the lab instruction's requirements, I identified water features in the Landsat 5 data. The image above contains a lake (the black feature). I used band combination 7, 4, 5 (RGB) that is used to delineate water bodies to make the lake stand out. I found it by viewing the histogram and determining what end of the histogram the spike was on and how large it is (to determine the brightness/size). I then examined Layer 4 in grayscale to find a large dark feature. I confirmed the feature pixel values fell in the 12-18 range using the inquire tool. 

    Snow features matched the lab instruction's requirements for the second feature. My layout image contains snow-capped mountains. I used TM False Natural Color band combination to make the snow stand out. I found it by viewing layers’ 1-4 histograms and determining what end of the histogram the spikes were on and how large it is (to determine the brightness/size). I repeated the process for layers 5-6. I then compared Layer 1 and layer 6 in grayscale to find a feature that matched the requirements. I confirmed the feature pixel values were around 200 in layers 1-4 and 9-11 in layers 5-6 using the inquire tool. 

    I identified a bay water feature that matched the image properties in the instructions. I used TM False Natural Color band combination to make the bay water stand out. I found it by comparing layers 1-3 and layer 5-6 in grayscale to find a feature that matched the requirements. The bay is brighter than normal in layers 1-3 and as dark as other water in layers 5-6.