Remote sensing Lab 6

Ethan Nauman

11/18/14

The goal of this lab is to introduce us to a very important processing exercise known as geometric correction. This lab was setup to develop my skills on two major types of geometric correction that are performed on satellite images. Technically this was a short lab but it also has parts of it that were tedious, especially when dealing with ground control points and repositioning them. 

Part 1: Image to map rectification

The skills that I used in this lab were performed in the Erdas viewer. I began with a blank viewer and uploaded the Chicago_drg.img image from our lab 6 folder. This image was a USGS 7.5 minute digital fast graphic (drg). I then opened a second viewer and uploaded the Chicago_2000.img image and ft both of the images to frame. I clicked on multispectral at the top of Erdas to activate these tools, then clicked on the control points tool handle. This opened the set geometric model window for me. Under this model I scrolled down and clicked on polynomial and then clicked ok. I used a first order polynomial equation in this window. When clicking on the select geometric model it opened two tools. The first tool was the multipoint geometric correction tool which I used as a guide in the collection and evaluation of the GCPs. The second tool was the GCP reference tool setup. I accepted the default parameters on the image layer (new viewer) on the GCP reference tool setup and clicked ok. I then navigated to our lab 6 folder and added our reference drg image- Chicago_drg.img. A window then opened that illustrated our reference system for the image. I accepted the default references and clicked ok. This process then opened the polynomial model properties (no file) window. At the start it said that there was no solution at the bottom of this window, however that would change after I started entering GCPs. I accepted the default parameters in this window and clicked on the close button. I then maximized the multipoint geometric correction window for better viewing purposes. This window contained two panes, one for the input image and another for reference image which was on the right. There was also two panes on the top of the window which were the zoomed in portion of the bottom panes. Once this window was open I had to delete the original GCPs that came with the image. To do this I selected all the point #'s in the bottom portion of this window and deleted them. The next step was to re-add four GCPs. I fit to frame both of the images in their respective panes and clicked on the crosshair tool that would allow for me to add GCPs. I started by adding a GCP on the input image then added the same point on the reference image as well. I repeated this process to add up to four GCPs. Once I added the fourth GCP my solution changed from no solution to solution is current. The next step was to evaluate the location of my GCPs by using the RMS error. I could individually see the RMS error on each GCP and also take the total RMS error in to make sure I was doing this correctly. To start my RMS errors were very high, the ideal situation would be to get them under 0.5, but since this was the first time dealing with GCPs we had to get them under 2.0. To get the RMS error below a 2.0 I had to zoom into the areas of the GCPs individually and relocate them while watching the RMS error and get it below a 2.0. This was a tedious process but it had to be done correctly. After getting my GCPs repositioned and the RMS error below a 2.0, my window appeared like this. 

Once my RMS error was below the 2.0 mark I was ready to run the geometric correction. I used the compute transformation matrix which had already been computed from my GCPs. On the multipoint geometric correction toolbar I clicked on the display resample image dialog button. this opened a window for me to save the image into my personal lab 6 folder. I left all parameters as their default and ran the tool. After running the tool my output image appeared as this. 

Part 2: Image to image registration

I started with a blank viewer in Erdas again. I displayed the Sierra Leone 1991 image from our lab 6 folder. This image had serious distortion, so i uploaded the Sierra Leone east 1991 grf image onto the original image. To actually see the distortion I used the swipe tool bar function. I right clicked on the images and scrolled down to the swipe function, this opened the swipe toolbar. I moved the slider around to see how bad the distortion actually was. After I was finished viewing the distortion I closed the swipe toolbar. I then cleared both the images and brought them into separate viewers like I did in the first part of this lab. My task was to correct the image just like I did in the first part of this lab with the GCP points. I clicked on multispectral to activate the raster tools and then clicked on control points like I did in the first part. I clicked on polynomial under the select geometric model  in the set geometric model window. I also clicked ok on the collect reference points forum in the GCP tool reference setup. I navigated to the lab 6 folder and added the reference image Sierra Leone east 1991 grf. I clicked ok on the reference map information window. On the polynomial model properties window I changed the polynomial order from 1 to 3 then clicked close. Just like I performed in the first part of the lab I deleted the original GCPs that came with the image and put my own GCPs onto the input and reference image. I ended with 12 GCPs when only 10 were required for a 3rd order polynomial. Putting and rearranging the 12 GCPs while maintaing a RMS below a 2.0 took awhile but I was able to get it done. After getting my total RMS error below a 2.0 my window appeared like this. 

After getting the RMS error below the 2.0 that was required I ran the model just like I did in the first part of the lab and my final maps distortion came out. The distortion was minimal especially because I used the 3rd order polynomial which meant I used more GCPs then the first part of this lab.