Background and Goals:
The main goal of lab 6 was to introduce and become familiarized with the geometric correction process associated with geometric correction. This assignment then was consistent of two parts with separate objects centered around that goal. They are:
Methods:
The first step in this assignment was to address data management. This meant that a new work folder needed to be created for the use of an active workspace for any data outputs that would be processed. This the case of this assignment, this folder was named "Lab6". Also, all data files for this assignment were previously provided by the instructor.
For both parts of this lab, the associated methods will be listed underneath each header.
Part 1: Map to Image Rectification
For the first part of this assignment, two images were loaded into two views within Erdas Imagine. The first was an uncorrected Landstat TM image of Chicago Metropolitan area. While the second image is a 7.5 minute map raster image of the same area. Once both images are uploaded intot he viewers, the fit to frame button was clicked so both images could be viewed side to side.
Within the "multispectral" toolbar, a control points button was accessed. This is the first step for geometric correction. Once this button is clicked, it opens up a "Set Geometric Model" dialog box. Within this box the parameter was set to polynomial. Once this is done, the ground control reference set-up box appears. All defaults were accepted and the 7.5 minute map raster was as the reference image.
From this point, four GCPs were collected. To do so, the first objective was to delete any previously listed GCPs. This was done to make sure no data errors occurred. Then the create GCP button was activated and a reference point was collected on the Landstat TM image. Then its corresponding location on the reference image was also collected. This process was repeated 3 more times. Once enough GCPs are collected, the image will show that the "model is current", this means that 1st order polynomial transformation can now occur. However, for geometric correction to be relatively accurate, the root square mean error (RMS) of the GCPs must be lower than 2. To make the RMS lower than 2, the GCP's are moved around to obtain closer proximity. This can be seen as figure 1 in the results section.
From here, the transformation matrix can be computed. To do so, the "Display Resample Image" tool was used. Within the dialog window, the output was saved to "Lab6" and all default parameters were kept. Figure 2 is this output image and is featured in the results section down below.
Part two: Image to Image Rectification
For the second half of this assignment, the process for image to image rectification was much of the same as the process for map to image rectification. However, some points differ. The first difference was the images that were brought into Erdas. The first image was an uncorrected Landsat TM image of Sierra Leone, the second image is a corrected Landsat TM image of the same location. This second image was used as a reference image for the first.
From here the same steps were followed as in part 1, until the "Set Geometric Model" dialog box was accessed, within the polynomial parameter, the order was set to 3. This means that 10 GCPs were needed to make the image show that the "model is current". However, for this geometric correction, 12 GCPs were collected. Once this was done, the resulting RMS needed to be below 1. Figure 3 shows this process within the results below. The final difference between the two steps was that when the image was resampled the resampling method used was bilinear interpolation. The resulting output can be viewed in figure 4, within the results below.
Results:
Below are the resulting images from part one and part two.
The main goal of lab 6 was to introduce and become familiarized with the geometric correction process associated with geometric correction. This assignment then was consistent of two parts with separate objects centered around that goal. They are:
- Part 1: To use a 7.5 minute map raster as a reference for Landstat TM satellite data of the Chicago Metropolitan area, in order to collect different ground control points (GCPs) that would be used for geometric correction.
- Part 2: To use previously corrected Landstat TM imagery of Sierra Leone as a reference for uncorrected Landstat TM data of that same location, in order to collect different GCPs that would be used for geometric correction.
Methods:
The first step in this assignment was to address data management. This meant that a new work folder needed to be created for the use of an active workspace for any data outputs that would be processed. This the case of this assignment, this folder was named "Lab6". Also, all data files for this assignment were previously provided by the instructor.
For both parts of this lab, the associated methods will be listed underneath each header.
Part 1: Map to Image Rectification
For the first part of this assignment, two images were loaded into two views within Erdas Imagine. The first was an uncorrected Landstat TM image of Chicago Metropolitan area. While the second image is a 7.5 minute map raster image of the same area. Once both images are uploaded intot he viewers, the fit to frame button was clicked so both images could be viewed side to side.
Within the "multispectral" toolbar, a control points button was accessed. This is the first step for geometric correction. Once this button is clicked, it opens up a "Set Geometric Model" dialog box. Within this box the parameter was set to polynomial. Once this is done, the ground control reference set-up box appears. All defaults were accepted and the 7.5 minute map raster was as the reference image.
From this point, four GCPs were collected. To do so, the first objective was to delete any previously listed GCPs. This was done to make sure no data errors occurred. Then the create GCP button was activated and a reference point was collected on the Landstat TM image. Then its corresponding location on the reference image was also collected. This process was repeated 3 more times. Once enough GCPs are collected, the image will show that the "model is current", this means that 1st order polynomial transformation can now occur. However, for geometric correction to be relatively accurate, the root square mean error (RMS) of the GCPs must be lower than 2. To make the RMS lower than 2, the GCP's are moved around to obtain closer proximity. This can be seen as figure 1 in the results section.
From here, the transformation matrix can be computed. To do so, the "Display Resample Image" tool was used. Within the dialog window, the output was saved to "Lab6" and all default parameters were kept. Figure 2 is this output image and is featured in the results section down below.
Part two: Image to Image Rectification
For the second half of this assignment, the process for image to image rectification was much of the same as the process for map to image rectification. However, some points differ. The first difference was the images that were brought into Erdas. The first image was an uncorrected Landsat TM image of Sierra Leone, the second image is a corrected Landsat TM image of the same location. This second image was used as a reference image for the first.
From here the same steps were followed as in part 1, until the "Set Geometric Model" dialog box was accessed, within the polynomial parameter, the order was set to 3. This means that 10 GCPs were needed to make the image show that the "model is current". However, for this geometric correction, 12 GCPs were collected. Once this was done, the resulting RMS needed to be below 1. Figure 3 shows this process within the results below. The final difference between the two steps was that when the image was resampled the resampling method used was bilinear interpolation. The resulting output can be viewed in figure 4, within the results below.
Results:
Below are the resulting images from part one and part two.
 |
| Figure 1: The top images depict the two images, and the four GCPs used in part one of this assignment. As is shown on the lower image, the RMS total is 1.8138 which is less than the two required. |
 |
| Figure 2: This image is the resulting output of the Landstat TM data after geometric correction. |
 | |
|
 |
| Figure 4: This image is the resulting output within part 2 of the Landstat TM data, after geometric correction. |
Comments
Post a Comment