This lab gives you the opportunity to explore a Landsat 4  7-channel Thematic Mapper (TM)  image. You can find more about the Thematic Mapper (TM) scanner under the Landsat Program particularly at Thematic Mapper resources here or here.  The images to be used in the exercise can be found in this compressed file that you can download and unzip.  You should place the unzipped folder on your flash drive and set this as your Main Working Folder - don't try to work from the desktop. If you must use the local hard drive copy the folder to the C-drive.  The study area this week is at the junction of the Little Colorado River and the Colorado River, date of acquisition 08/24/85.  You can find a topographic map of the image area here.   A photo of this location can be found here and a description of the river can be found here.  Explore these images following the instructions below and answer the questions at the end of the document to complete the exercise.  You will also produce a color composite image at the end.  This lab activity may be done with a partner.

1. Open Idrisi , create a new project, and set the environment to the folder that you have copied from the downloaded zip folder above to your flash drive or to the C-drive. Open all seven channels of the image using the display launcher. You may try these either with or without autoscaling. A couple of words about autoscaling:

Autoscaling is the automatic division of a range of data values into a new range of values for display. In the IDRISI32 DISPLAY system, autoscaling is used to display images with any range of values with the number of colors in a specified palette. Autoscaling performs a linear stretch of the data values (i.e., it assigns the lowest data value to the lowest palette index number, the highest data value to the highest palette index number, and all other data values to a palette index in direct proportion to their position within the data range). In IDRISI32, autoscaling is performed on the fly and does not create a new file. A new file identical to that displayed with autoscaling may be produced by using STRETCH with the Linear option. Autoscaling is also used in DATABASE WORKSHOP to scale the attributes in a field to symbol file indices (0-255) for display.

Use the grey256 palette for clarity. Answer question 1 and 2 below.

2.  Use the system tools to produce image histograms of the COLO1 and COLO7 images.   You can either use the histogram button on the LAYER PROPERTIES dialog or you can use the HISTO button on the speed bar.  Using the HISTO button gives you more control over the parameters of the graphic histogram.  A few words about histograms:

HISTO first requires the name of the input image to be analyzed. You are then required to either input a desired class width (the default is set to 1) or determine the number of classes into which the values should fall.

When you enter the name of the image to be analyzed, HISTO automatically presents the minimum and maximum values in the image. They can be reset for the purpose of creating the histogram only, but note that the image documentation file will not be changed. Use this opportunity to set the beginning of the first histogram class and the end of the last. It is advantageous to reset the minimum to zero whenever possible since a simple relationship then exists between histogram classes and data values (particularly useful with the numeric option). Set the maximum to exceed the largest value in the file since the histogram maximum excludes all values greater than or equal to the value specified.

Finally, HISTO requires that you specify whether graphic or numeric output is desired. The graphic output is the default. If you are uncertain about the data ranges of these classes, run HISTO a second time, and choose numeric output. Numeric output specifies the data ranges for each class. Numeric output also lists the frequency of cells falling within the class, the proportion that this represents of all cells and the cumulative frequency and proportion -- i.e., the frequency and proportion of all cells in that class and all lower classes.

Next to the graphic display and at the end of the numeric data output, summary statistics are given that include the mean, range and standard deviation of the image. Both output types list the class width, maximum, minimum and mean of the histogram, and actual maximum and minimum. Classes below the histogram minimum and classes above the histogram maximum are not shown.

The graphic ouput allows you to change the format of the histogram between bar, line and area graphs and between non-cumulative and cumulative modes. You may save the graph as a .bmp, .wmf or .emf file by pressing the Save Graph As button. The current graph may be printed by pressing the Print button.

The numeric HISTO output may be printed, copied to the clipboard or saved to an ascii file by pressing the corresponding buttons.

 

Compare these two histograms and answer question 3 below.  You should use the "Copy to Clipboard" option to copy the histograms and then paste them into your word document to illustrate your answer to question 3 below.  Remember to use all the tools you have available to you for exploring the images (e.g. cursor inquiry mode, window, etc.).

3.  Use the REGRESS function (see under menu ANALYSIS/STATISTICS) to populate a CORRELATION MATRIX for the seven channels in your lab folder.  This will require you to run the REGRESS module for every possible two-variable combination of the seven channels (How many times would this be?).  The REGRESS module looks like this:

A few words about the REGRESS module:

Regression is a statistical technique that allows one to examine the relationship between two quantitative variables. This relationship is expressed in terms of the correlation between the variables (i.e., their degree of association) and a best-fit trend line that expresses mathematically the character of the relationship. IDRISI provides a simple linear regression procedure with the module named REGRESS.

REGRESS first requires that you specify the type of regression, either between values files or image files. You then must input the names of the independent and dependent variables. In addition, if image files are being regressed, you have the option to use a mask file to limit consideration to only a subset of pixels. A mask image is a binary integer or byte image with non-zero values in all cells that should be considered.

A graphic scatter plot and trend line are displayed with the regression equation, correlation coefficient, degrees of freedom and t-statistic. If you want a printout, just click on the printer icon. If you want to copy the results of the regression to the clipboard, click on the copy to clipboard icon.

To create a correlation matrix modify your Word document to include a table that looks something like this:

channel	1	2	3	4	5	6	7
1
2
3
4
5
6
7

... or you can print this table.  Note that an image channel correlated to itself will produce a correlation coefficient (R) value of 1.  Write or type the R values for each 2-channel combination into the table.  These will be used in #4 below.

4.  Use the COLOR COMPOSITE function under the DISPLAY menu (or use the button) to produce at least three (3) different color composite images for this scene. Use the CORRELATION MATRIX above to determine which channels to use for your composites.   Use three different channels with the LOWEST correlation values. Make sure you use the setting for Linear with saturation points when making the composites. A few words about Color Composites from Idrisi help:

Use at least four (4) different band combinations (although you should feel free to use more).

Questions - Lab 2

1. There is a striking visual difference in this image of the Colorado and Little Colorado Rivers, particularly the way that these two rivers appear visually, between channels 1-3, and channels 4,5, and 7. 
   
   a. As an image interpreter how would you describe the difference between the two rivers in a visual sense?  Remember SSPTHTSSAR...
   b. How would you describe it in a digital or numerical sense? (hint: you should use the tools in Idrisi to explore and sample the image for reflectance values in particular pixels and develop a sense of what this difference is numerically across the spectral range of the TM sensor...)
   c. How would you describe it in a radiometric sense? (hint: try to describe it in a generic non-numeric sense across the spectral range of the TM sensor...)
   d. What would you hypothesize as the cause or causes for this difference in the way the two rivers appear?
    
2. The image COLO6 is a TM channel 6 (thermal) image that senses the temperature of the surface being imaged.  Would you rather have your canoe tip over in the Colorado or the Little Colorado River if you are not wearing a wetsuit?   Why?
    
3. Compare the two histograms and describe what you see, paying particular attention to the minimum and maximum values, the distribution of the values across the range, and particular local minima and maxima.  How would you explain the difference in the two histograms?  Is there a recognizable problem with the histograms as you see them?
    
4. Which three two-channel combinations have the lowest correlation coefficients (R)?   Which three have the highest?  Why do you think this is.
    
5. Print your best color composite using the Designjet plotter formatted as a C-size composition.  Write a short paragraph describing the technical details of its production and any significant observations that you can make about the display and turn the answer to these questions and your image in to signify the completion of the lab activity.  Aces Again!!! This activity is due at the start of class April 28.