GEOG 338: Lab 6

Lab 6: Geometric Correction

Melissa Hackenmueller

Background

Lab six was designed to introduce students to geometric correction. Geometric correction is very important in image pre-processing for it needs to be used to accurately extract biophysical and sociocultural information. Two types of geometric corrections will be performed in this lab. The first is image-to-map rectification and the second is image-to-image rectification.

Methodology

            The first section of this lab uses a USGS 7.5 minute digital raster graphic image of the Chicago Metropolitan Area and a corrected Landsat TM image of the same area. The first step is to input both the Chicago_drg.img and the Chicago_2000.img into Erdas Imagine. Next, I clicked on the control points tool of the multispectral toolbox. A select geometric model dialog will open. I selected polynomial, then image layer, added my reference map (Chicago_drg.img), and then hit okay one last time in the reference map information dialog. This then will bring up a multipoint geometric correction window. This window showed my input image (Chicago_2000.img) on the left and my reference image (Chicago_drg.img) on the right. Ground control points could now be added to my maps, but I first made sure that any existing ground control points were deleted. Then I used the create ground control points tool to input a point on each map in the same area. I then repeated this process three more times to get four ground control points that were evenly distributed across my images. Once, this was done I looked at my total RMS error and it was very high. To decrease the RMS error, I zoomed into each ground control point and adjusted it to accurately represent the area on the other image. Once my RMS error was below 2 I decided that my image was spatially correct enough for this project (Figure 1). Ideally the RMS error should be below 0.5 for the geometric correction to be accepted in the remote sensing field. I was content with my RMS error so I used the display resample dialog button to create a rectified, geometrically correct image.

            Part two of this lab was to use image-to-image rectification to create a geometrically correct image of the east side of the Sierra Leaone. I started by importing the sierra_leone_east1991.img and the sl_reference_image.img into two separate viewers in Erdas Imagine. I then chose control points from the multispectral toolbar. In the geometric model window I selected polynomial, then image layer, imported my reference map (sl_reference_image.img), accepted the default inputs for the reference map information, and finally chose 3^rd order polynomial. I then clicked on the create ground control points tool and started to add my points as I did in part one. This time I added 12 ground control points to my image. Only 10 are needed for 3^rd order polynomial, but a few extra points allow the rectification to be more accurate. Once I placed all my ground control points, I more accurately placed them until my RMS error was below 1 (0.5 would be ideal, but was not necessary for this lab). Figure 2 is an image of my RMS error. I was content with my RMS error so I used the display resample image dialog button to create a geometrically correct image (sl_east_gcc.img). This image could then be used to more accurately analysis the area of interest.

Results

Figure 1: This image displays the ground control points that I used for my image-to-map rectification.
The ground control points are distributed evenly about the image to provide an accurate output image.
My RMS error is below 2, which was the goal for part one of this lab.

Figure 2: This image displays my ground control points for my image-to-image rectification. 

The RMS error is below 1 as that was the goal for part two of this lab. 

References

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.