Lab 6: Geometric Correction

Introduction

The goal of this lab is to familiarize students with an image pre-processing technique known as as geometric correction. Students are able to exercise this skill using two different types of rectification pre-processing: image-to-map and image-to-image rectification.

Methods

For both sections, students use the Control Points tool under the Multispectral toolset to begin the geometric correction process. Using a Polynomial Geometric Model, Erdas users have the option to choose the number of polynomials they wish to model in their rectification transformation. The number of polynomials required for use may vary based on the extent of the distortion between the original and referenced image. 

In Part One, the distortion that exists between the original image and the referenced map is minimal. Students use a 1st order polynomial equation to rectify the desired image. A 1st order polynomial equation requires a minimum of three Ground Control Points (GCPs) between the map and the image in order to make the transformation. For the purposes of this lab, four were placed for good measure. 

Figure 1: Image-to-Map Rectification Process using a 1st Order Polynomial Equation

For the image-to-image rectification process in Part Two, the distortion that existed between the two images was more drastic. Students utilized a 3rd order polynomial equation to execute the geometric correction process, requiring a minimum use of 10 GCPs between the two images. Again, to increase the accuracy in the transformation process, students added an additional two points above the minimum requirement. 

Figure 2: Image-to-Image Rectification Process using a 3rd Order Polynomial Equation

For both Parts One and Two, once the original GCPs were placed, students would revisit their locations, zooming in closer each time, and ensuring that they were placed in the correct locations on the original image as the referenced layer. Small adjustments were made each time until the Root Mean Square (RMS) error in the lower right corner of the display symbolized a value less than or equal to 1. 

Results 

The rectified image in Part One is pictured to the left of the original photograph in Figure 3 below. In it, one is able to see that the image rectified image proves to be much less grainy in detail and contains much smoother details in its landscape transitions. The resulting photograph utilized the nearest neighbor resampling technique to configure its final results. Thus, students have demonstrated the execution of an intensity interpolation in the figure below. 

Figure 3: Results of the Image Rectification Process from Part One

Figure 4: Overlay of Rectified Image to the Referenced Image using the Swipe tool

The geometrically corrected image from Part Two of the lab resulted in a geometrically accurate, nearly perfect overlay of the resulting rectified image over the referenced photograph, as can be seen in Figure 4. For this rectification, the bilinear interpolation resampling method was utilized, as it provides more spatially accurate results.

Conclusions:

This exercise allowed students to experiment with two techniques of Geometric Correction and image rectification. Geometric Correction of photographs is an important skill to learn because it is a common pre-processing technique that is utilized in Remote Sensing and for photo interpretation. 

Sources:

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.