The purpose of Geometric Corrections lab is to develop skills using two major types of geometric corrections. These corrections are often performed on satellite images before processing in order to better the data quality and integrity. This process helps to align aerial images. Aerial images are rarely if ever perfectly inline due to factors such as the differences in altitude or the angles of the images. Rectification of an image is the process of changing a data file coordinate to a different coordinate system known as a reference system. The two forms of geometric correction are listed below.
1. Image-to-Map Rectification: Map coordinates systems are used to rectify the image data to the correct pixel coordinates.
2. Image-to-Image Rectification: Previously corrected images of the same locations are used to rectify the image data pixel coordinates.
Methods
The first method that was used was the image-to-map rectification. The Chicago_drg.img was brought into viewer one and fit to frame, this is a USGS 7.5 minute raster graphic (DRG), that covers part of the Chicago region and also adjacent areas. A second viewer was then opened and Chicago_2000.img was opened there.
Under the multispectral tab in the top right, control points was selected. Under the Select Geometric Model tab the polynomial box was checked. By selecting the geometric model, two tools were opened, the multipoint geometric correction tool and the GCP tool reference setup. All of the default settings were accepted in the new viewer. The Chicago_drg.img was then brought in from the lab 6 folder that was previously copied over into my own personal folder in the Q drive. In the reference map information click okay, from there the polynomial model properties was displayed and before the addition of GCP's it reads model has no solution. Make sure to keep both images at full extent, as the software will crash repeatedly if that is not done.Before additional GCP's can be added the previous ones are deleted, this is done by highlighting the GCP's and right clicking and selecting delete selection.
Next, three pairs of GCP's were added on the images, this was done by clicking on the Create GCP tool. Once three are added the image will now read "model solution is current," and now GCP's can be added by clicking on only one of the images. Look at the root mean square (RMS) error to see how accurate the GCP's are. For this part of the lab the RMS error should be under 2. The GCP's can be moved by zooming in and moving the GCP until the RMS error in the table on the bottom gets under 2. This process was repeated for all the GCP's. A screen capture is provided below in figure 1 that also shows the table, showing the RMS error under 2.
From here, the display resample image was clicked and the output image was rename to Chicago_2000gcr.img and saved in the folder in the Q drive. All of the default parameters were accepted and the image was then brought into a viewer to view the improvements.
 |
| Figure 1 |
In part 2 of this lab the majority of the steps are repeated from part 1. Image sierra_leone_east1991.img was brought in and fit to frame and the second image sl_reference_image.img was brought into the second viewer. The swipe function was activated to see the extreme distortion in the images. From there follow the steps described above to get to the point of inserting GCP's. All the same steps were followed for all GCP's added. The Display Resample Image button was clicked again and the image was saved as sl_east_gcc.img in the lab 6 folder in the q drive. The resample method was changed to bilinear interpolation and all other defaults were accepted. This processing takes some real time, so be prepared to wait. The corrected image was then brought into Edras to take note of how much better the quality is.
 |
| Figure 2 |
Results
This lab helped to give a good basic skill set in Geometric Correction. Making sure that images are correctly rectified is essential to putting out high quality and accurate images for analysis. It is remarkable how an image can appear to be correct in comparison to another one, but looking at the root mean square error can show something different. The ground control point locations this lab were previously selected by Professor Wilson, though selecting good locations for GCP' is essential to the process.
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.
No comments:
Post a Comment