LAB 6: Geometric Correction

INTRODUCTION

The following lab demonstrates the ability to perform geometric correction. This is an important activity used in satellite image preprocessing before interpreting and extracting data. Both a 1st order and a 3rd order polynomial transformation was performed. 

METHODOLOGY

For the first order polynomial transformation, a Landsat TM image of the Chicago metropolitan Area was corrected using a USGS digital raster graphic of the same area. Using ERDAS Imagine, both images were displayed side-by-side in 2d-viewers. The process began by selecting a polynomial geometric model from control points. The input image was the Landsat TM and the reference image used was the USGS image. Four ground control points were then placed in the same location on both images, all spread away from each other to produce an accurate geometric correction for areas on the imagery. A careful examination of spatial accuracy of matching GCPs was made by maintaining a root mean square (RMS) error of less than 0.05 and a control point error total of less than 0.5.

For the third order polynomial transformation, a corrected Landsat TM image of the Eastern Sierra Leone region was used to correct satellite imagery of the same region. The same process was executed, except a third-order polynomial transformation was performed. A total of 12 GCPs were matched and the RMS error and total control point error was maintained at less than 1.

RESULTS

Figure 1 shows the matching four ground control points located at nearly the same exact location in both the input and reference image. This is only a first-order polynomial transformation, so only three matching GCPs are required. 

Figure 1

Figure 2 shows the same view as figure 1 except for the Eastern Sierra Leone region. Since a third-order polynomial transformation is performed, a total of 10 GCPs is required.

Figure 2

Figure 3 shows the geometrically-corrected image over the pre-existing Landsat TM imagery. The grayer image is the output image and the correction can be seen by the curving of the left side of the image and the abrupt change on the right edge where the cutoff point is between images. 

Figure 3

SOURCES

Satellite images were obtained from Earth Resources Observation and Science Center, United Sates Geological Survey.
Digital raster graphic was obtained from Illinois Geospatial Data Clearing House.