Winter 2018 GEOG457/657: Lab 1 – Review of Intro to Remote Sensing
Lab as an assignment
In this lab, you are going to complete your first assignment for the course – an easy one…
We are going to work with Landsat TM data – most likely from Landsat 4-5 – to complete a series of tasks that illustrate your knowledge of Remote Sensing and the use of PCI Geomatics. As the course progresses, Scott will introduce the use of other (Open Source as much as possible of course) software in performing the same tasks that we use PCI.
1.) Obtain a scene from Earth Explorer that is low in cloud coverage (at least 10% of less). Bring it into PCI and create your own PCI file (pix file) in the what Roger and Scott would call the ‘best practices’ way of doing so. HINT: You may want to clip down the scene to not only make it smaller, but contain the elements you will find easy to classify later on.
2.) Review the data to determine which bands should be used for input into a classification.
3.) Perform an unsupervised classification on the dataset with a reasonable set of classes.
4.) Create two ratios from the data set:
- one for NIR and Visible Red
- one for MIR and Visible Green
5.) Using the same dataset, create an NDVI and Tassel Cap transformation layer(s).
6.) Run the unsupervised classification again but include the two ratios, NDVI and Tassel Cap in different combinations until you are most happy with your classification. For instance, you may find the Tassel Cap channels on their own as the most useful for classification.
7.) Compare the results of these classifications, Use the sieve algorithm on the data to eventually produce a Vector dataset representing the different classes in the scene. You are welcome to create vector data for all classifications, but only the classification you are most happy with is required.
7.) One the same scene, perform a supervised classification with the same classes you brought out from the unsupervised classification. Load the vector layers your created from the sieved results into you projects and see how the seeding results compare with the polygons from the unsupervised classification.
8.) Sieve out the results from the supervised classification and compare the results to the unsupervised.
9.) Write up the results by including the resultant images into a word (or whatever you prefer) document.
10.) Write up:
- Jot down the steps you took to complete each of the 8 steps in the assignment (a mixture of point and prose form)
I chose bands X,Y,Z because of the low correlation between them for my input for my classification. The band correlations that let to this decision are:
- band 1 vs band2 – correlation coefficient of .90
- band 2 vs band 3 – correlation coefficient of .88
Things to remember