Remote Sensing Lab 6: Geometric Correction

   Goals and Background: This lab was meant to familiarize me with geometric correction of images used in preprocessing before images can be studied for spatial-statistical information about observed phenomena. We used a USGS 7.5 minute raster of chicago in order to correct a Landsat TM image of the same area (image to map rectification), and we used a Landsat TM image of Sierra Leone to correct a distorted image of the same area (image to image registration).

   Methods:

Part 1 (Image to Map Rectification): Starting with a new 2D viewer in ERDAS Imagine, I opened the USGS Chicago DRG reference file. I then opened a new view and displayed the Image of the same area I intended to geometrically correct. I then made sure that the view of the image I wished to correct was highlighted. Next, I clicked on multispectral, then control points, opening the Set Geometric Model dialog. I now clicked on Polynomial, then okay, this correction being done with a first operation polynomial process. Next I clicked on image layer (new view), and okay. Navigating to the my personal folder I added the USGS image which I wished to reference, and clicked okay on the reference map information dialog. Going on, I accepted default model properties, and maximized the window in order to have more space to do the sensitive work of finding and moving matching points on the maps. I now deleted the current GCPs and created three new ones, selecting precisely the same place on the two images, then moving one of them while zoomed in extremely far while watching my RMS value decrease. After my total RMS was low enough for my standards, I clicked the display resample image dialog button, after which I specified my output location and specific name and accepted he other default parameters and clicked okay.

Part 2 (Image to Image Registration): Bringing both the image I wanted to correct and the reference image into a single viewer in ERDAS Imagine, I right clicked in the viewer and clicked swipe. I now used to slider to observe the difference between the images (the extent to which the one image was distorted). I now closed the viewer swipe window, then cleared the reference image from the viewer. I now clicked on multispectral, then control points, selecting polynomial, and setting my reference image file in the resulting windows before the main interface window. On one of the windows I observed the coordinate system the image was in, and on another I changed the polynomial order to 3. I now deleted the GCPs already present, and created 10 of my own, correcting as need be in order to get an adequate RMS before adding more GCPs automatically placed for good measure and to prevent the wrap tool error. I now saved and named the file and changed the resample method to bilinear interpolation.

   Results: 

Corrected Chicago

Corrected Sierra Leone

   Sources:
Data was supplied by my instructor. Satellite images are from Earth Resources Observation and Science Center, United States Geological Survey. Digital raster graphic (DRG) is from Illinois Geospatial Data Clearing House.