Scribd
Upload
299 views

Lab 3 Classification

This lab assignment involves classifying land cover types in an ASTER satellite image of Yuma, Arizona using different classification methods. It also involves applying atmospheric corrections, including dark object subtraction and MODTRAN transmissivity correction, and analyzing band ratios to reduce topographic effects.

Uploaded by

Kaleab Tekle
Copyright
© © All Rights Reserved
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
299 views

Lab 3 Classification

This lab assignment involves classifying land cover types in an ASTER satellite image of Yuma, Arizona using different classification methods. It also involves applying atmospheric corrections, including dark object subtraction and MODTRAN transmissivity correction, and analyzing band ratios to reduce topographic effects.

Uploaded by

Kaleab Tekle
Copyright
© © All Rights Reserved
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 3

ESS422 Name:_________________________

This lab is due Thursday, April 23rd.

LAB 3: Classification

In this lab you will examine different classifications, apply an atmospheric correction and try channel
ratios. The image is the ASTER image of Yuma, Arizona with channels 1, 2, and 3 (VNIR: Green, Red,
Near IR) re-sampled to match channels 4-9 (SWIR) in size and pixel resolution.

A. Open Yuma_ASTER.HDF

1. Choose several types of landscapes (e.g., forest, water, etc.) from the scene using the ROI tool.
Run four classifications: K-Means, Parallelepiped, Maximum Likelihood and one of your
choice.

On the main menu click on Classification> Supervised


Choose Parallelepiped or Maximum Likelihood
In the Parameters Box choose 'Select All Items' for your regions
Output result to Memory or choose an output file name
Change 'Output rule image' to 'No'
Click OK
Feel free to experiment with the Standard Deviation Parameters
K-Means will be in Classification> Unsupervised

a. Compare the results of the four different classifications. Note your observations.

b. Which classification do you think was the most successful, and why?

B. Apply a 'dark object subtraction' atmospheric correction to the image. This is the Path Radiance
correction.

1. Find a hill in the image that looks to be the same except for lighting geometry. Record the
DN's for channels 1 and 3. and their ratio value for the sunny and shaded side. It the ratio the
same or different on sunlit and shaded sides of the same hill?

2. Find the DN value of the darkest pixel in the image. Make an ROI of a group of pixels
surrounding the darkest pixel (to be sure it's not an artifact or noise). Record the approximate
location of the region.
3. Use 'Dark Subtract' for the correction:

Go to Basic Tools > General Purpose Utilities> Dark Subtract


Click on the Yuma file, then OK
Choose Region of Interest as the Subtraction Method and select the ROI
Choose an output file name.

4. Remake the ratio of Step 1 now that you have corrected for atmospheric path radiance. What
are the ratio values now?

5. Sketch the histogram for channel 1 before and after the correction and note any changes.

C. Apply a transmissivity correction to the image using MODTRAN.

1. Use the PPT tutorial to set up and run MODTRAN.

2. Once you have the wavelength in microns, average the transmissivity values over the
nine Aster channel ranges. Record the averages below. An average is being used to
simplify this exercise.

Aster Channel Wavelength Ranges Average Transmissivity

Channel 1: 0.52-0.60 ___________________________

Channel 2: 0.63-0.69 ___________________________

Channel 3: 0.76-0.86 ___________________________

Channel 4: 1.6-1.7 ___________________________

Channel 5: 2.145-2.185 ___________________________

Channel 6: 2.185-2.225 ___________________________

Channel 7: 2.235-2.285 ___________________________

Channel 8: 2.295-2.365 ___________________________

Channel 9: 2.360-2.430 ___________________________


3. Apply the average Transmissivity correction to the image using 'Apply Gaim and Offset.'

Go to Basic Tools>General Purpose Utilities>Apply Gain and Offset


Choose the file that has been 'dark object' corrected, click OK
Under 'Gain Values' click on each band (channel) and enter the corresponding
transmissivity value. Offset values should be zero since the dark object subtraction
has already been made. Both steps can be performed together using this utility if you
do 'stats' on the dark ROI, then record the average in the 'Offset Values' option.

This will give the final atmospheric correction. View the images and their histograms and note
your observations.

4. If you were to remake the ratios from B again now that you have corrected for transmissivity,
would the new values be the same or different from these in B 4?

D. Band Ratios minimize surface roughness effects at all scales.

1. Using Band Math or Transform>Ratios, make three ratios of the atmospherically corrected
VNIR channels (1/2, 1/3, 2/3). Are the ratios useful in reducing topographic effects?

You might also like