Exercise #6 by ls723a4r


									Exercise #4                               Name: _______________________________
Geography 475
Digital Image Processing
Geometric Correction
Due: March 5, 2010

In this exercise, you will learn to rectify an image to a map, and to register other images
to a map-rectified image. In addition, you will subset the area of a county from geo-
corrected imagery. The image files have been atmospherically corrected and calibrated
so digital numbers (DNs) represent 8-bit surface reflectance (in percent). Both images
are small subsets of a full Landsat TM scene of the Traill County, ND, area. There are a
few different ways to conduct a geometric correction in ERDAS. You will look at one

Part I: Image Rectification

Copy all files from the Exercise 4 folder on the server to your local hard drive. Change
permissions to allow write access. Start Imagine and open the file
traill_county_050992.img. Enlarge the viewer and fit the image to the window. To rectify
an image, you need a map to use as reference. For this scale of imagery (TM at 30 x 30
m), USGS 1:24,000-scale topographic map sheets are appropriate. Often, it is
necessary to gather points manually off a paper map. This is a time-consuming process.
A less tedious task is to use digital map products available for download.

A. Digital Map Download and Import

There are many 1:24,000 USGS quads that cover the entire area of Traill County. A
good source of seamless DRGs is the USGS’s National Map Seamless Server
(http://seamless.usgs.gov/). I have downloaded, re-projected (from NAD27 UTM to
NAD83 UTM), and masked a seamless 1:24,000 DRG for Trail County on your behalf.

Open a new viewer and load the file traill_county_24k_drg_wgs84_masked.img.

B. GCP Selection and Transformation Calculation

You are now ready to start the rectification process. In the viewer displaying the Landsat
image data, select Raster | Geometric Correction. Set the Model to Polynomial and
click OK. Since you are dealing with Landsat data, leave “Polynomial Order” set to 1.

Next, click on the Projection Tab. This is where you must indicate the map projection
you want to transform the image data to. Since you want it to match the projection of the
DRG, select Set Projection from GCP tool. Then, select Existing Viewer and click
OK. Imagine asks you to left-click inside the viewer containing reference coordinates.
Click anywhere inside the viewer displaying the DRG. Complete Table 1.
                        Table 1: Output Projection Information

Click OK. Loaded are 2 “chip extraction” viewers (small viewers showing a zoomed in
area of the image) and the GCP tool. Locate the town of Hillsboro on both the image
and on the DRG.

The key to selecting GCPs is to find features that are readily identifiable on both image
and map. The best GCPs are road intersections. You must locate GCPs accurately! Pay
attention to detail! Find a road intersection on the image that you can identify on map
near Hillsboro.

In the GCP tool, click the create GCP icon:     . Use the mouse to “drop” a GCP on the
location of the road intersection on the image (left-click to place the GCP). Now look at
the GCP editor. Note that a file coordinate (X Input, Y Input) is entered for GCP #1.

Again click the create GCP icon and drop a GCP on the location of the road intersection
on the map. Now the GCP tool displays a map coordinate (X Ref., Y Ref.) for GCP #1.

A few hints: A) If you are having difficulty seeing the GCP, you can change its color by
right-clicking in the GCP tool under “Color” and selecting a desired color. You have to
do that for both the file and image GCP. B) If you are not completely satisfied with the
location of the GCP, you can fine-tune it by moving the point around in the chip
extraction viewer. C) If you wish to delete a GCP, left-click on the appropriate row in the
GCP tool under “Point #” and then right-click and select Delete Selection.

Repeat the process to collect at least 12 GCPs. Remember that GCPs should be
spaced widely (not clumped in the same geographic location). Note that after you collect
the third GCP, Imagine starts to calculate an RMS error and attempts to place reference
GCPs for you (which is a very nice feature).

After you have placed at least 12 GCPs, take a look at the total RMS error. It should be
less than 1. The lower the error the better.

1. What is your RMS Error?

If your error was less than one, skip this paragraph. If RMS is greater than one,
examine RMS for each individual GCP. Is there one that has high error? If so, delete it.
Continue to delete GCPs until the RMS is less than one. If you are down to 8 or 9 points
and error is still greater than 1, get help!
C. Image Resampling

O.K., the GCPs are selected and the transformation is calculated. Time to resample the
image. In the small Geo Correction Tools window, click on the Display Resample Image
Dialog icon: . Call the output file traill_county_050992_rectified.img and save it in
your directory. Leave “Resample Method” set at nearest neighbor. Check to make sure
the projection information is correct. Change “Output Cell Sizes” to 30 m. Click OK.

   2. What is the advantage of using the nearest neighbor resampling method?

   3. What are the disadvantages?

Load the resampled image into a viewer. Load the inquire cursor. Note that map
coordinates are now in UTM. Change the coordinates to Lat/Lon.

   4. Do latitude, longitude coordinates display?

   5. Do you see evidence of line breaks?

   6. What visual evidence is there that the image is now geometrically correct?

Part II: Image Registration

Close all viewers and windows except for the main Imagine tool bar. Do not save any of
the transformation information or geometric correction models.

Start two new viewers. Into one, load the file you just created
(traill_county_050992_rectified.img). Into the other, load traill_county_070388.img. In
the viewer displaying traill_county_070388.img, select Raster | Geometric Correction.
Select Polynomial and OK. Again use a 1st-order polynomial. Select the Projection
tab, and select Set Projection from GCP tool. Leave the selection set to Existing
Viewer, click OK, and click then anywhere inside the viewer displaying
traill_county_050992_rectified.img. When the projection information is displayed, click

Your job is much easier this time. All you have to do is match locations between the
uncorrected and corrected image. Use the GCP tool to find at least 10 matching points.

   7. What is your total RMS error?

If less than 1, resample the image (make sure to set the output pixel size to 30 m). Call
the output traill_county_070388_rectified.img and save in your directory. If RMS error is
greater than 1, check for individual GCPs with a lot of error.

After the image resamples, load it into a viewer.

A. Subset of a Specific Study Area

There is a file (ARC/INFO GRID) representing the area of Traill County in the exercise4
folder. You will subset each of the geo-corrected images by this file.

Close all viewers and windows except for the main Imagine tool bar. Click Interpreter |
Utilities | Mask. The input file is traill_county_050992_rectified.img. The mask is
trail_county_mask.img. Call the output file traill_county_only_050992.img and save in
your working directory. Click OK. After the process completes, load the new file into a
viewer. Repeat the process for traill_county_070388_rectified.img.

You now have radiometrically-correct, registered images for the area of Traill County
ready for analysis!

To top