|Home - Teaching - Research - About Us - Software|
We begin with the assumption that you already know how to open, dispaly various band combination, and produce X, Y, and Z profiles
1. Begin by displaying the georegistered and atmospheric corrected image (such as the Landsat image of the Olympics that is in ERDAS .img format - R:\lawers\data\landsat\8_27_98\oly.img). I've displayed a 543 band combination.
2. The supervised classification approach requires that we define training area. These area are defined as Regions of Interests (ROI's). Select the OVERLAY > Region of Interest option from the Image display menu.
As you can see the ROI can be captured from any of the display windows. The first ROI will be display in the color red. The shape of the ROI is up to you. It should define the extent of the pure surface material which you are interested in. CLICK and HOLD the left mouse button on the edge of the Region you wish to define and DRAG the mouse to outline the region. RIGHT click once to close the polygon and RIGHT click again to select or enter the ROI.
As long as you do not click on the NEW REGION button in the ROI tool window all other defined regions will become part of the same (or selected ) ROI. In this way, you should think of the ROI as a spectral region as well as a spatial region.
Define the ROI's for the classes you want. Use both the Image window and the Zoom window if you like.
3. We can also make training sites directly from the specta. Create a 2-D scatterplot of two band feature space (I suggest bands 4 and 3) - go to the Image window menu bar TOOLS > 2-D Scatter plot. Use your LEFT mouse button to CLICK and HOLD while drawing a polygon around the points in the scatter plot that are the spectra of interests then RIGHT CLICK and the points and pixels will be selected (they will be painted in the image). When you've defined the spectra you are interested in RIGHT CLICK and select EXPORT CLASS. The spectra will be added to the ROI list. You can edit the color if you like.
4. Once you've completed your training site (ROI) collection you can classify the image using this information. Remember that by using the training site options you are suggesting that you know before hand what the surface material is at these locations. So you can edit the class names now to identify that ROI with that material.
5. From the Main ENVI tool bar select CLASSIFICATION > SUPERVISED > PARALLELPIPED. Select the image you wish to classifiy. Select ALL of the ROI's you wish to us and define the location of both the output and the RULE files. Click OK. When completed display your results
Not very good. But then my examples were very limited in the variety I accepted in each of my ROI's.