Lab 6: Geometric Correction

Lab 6: Geometric Correction


Goal and Background

The goal of this lab is to utilize geometric correction, which is the process of removing geometric distortion in imagery to put individual pixels in their proper planimetric position. Two types of Geometric correction will be performed on two different images. The first is an image-to-map rectification, which uses a map to correct a satellite image. The second is an image-to-image rectification. this uses an already corrected image to rectify an image that is distorted. These processes are part of preprocessing activities which are performed before biophysical and sociocultural information can be extracted from satellite images.

Methods


Part 1: Image-to-Map Rectification

In the first part of this lab an image of the Chicago Illinois (Figure1) area was to be recorrected through the use of a topographic map (Figure 2) of the same area over the Chicago region of Illinois. This correction of the image was done using a First Order Polynomial Transformation with a nearest neighbor method. This order of transformation requires the use of a minimum of 3 GCP points. GCP points are ground control points which are used by placing one on the map, and the goal is to place an adjacent point on the image which is being recorrected. To place these GCP points, the control points raster tool on Erdas Imagine was used (Figure 3) and both images were brought in, and the process of adding GCP points is done here by lining up the images and finding the same point on both the image, and the map. In placing 3 points on the images, a point error is shown on the bottom right of the control points tool (Figure 4) which provides a degree of how close the GCP points are. The greater this error number is the less spatially accurate the image is. In order to have a spatially accurate image, the error number should be below 2 in the case of this image. After 3 points were placed a 4th was placed to increase the accuracy of the image. It is also important to spread out the GCP points over the image, so the geometric correction doesn't just focus on one small area.
 Figure 1: This is the uncorrected image of the chicago Illinois area.
 Figure 2: This is the topographic map that is used to rectify the satellite image of the chicago area.
Figure 3: This image shows the GCP points used between the map and the image, to rectify the image on the left.
Figure 4: This image shows the control point error number which is located on the bottom right of the control point tool in Erdas Imagine.

Part 2: Image-to-Image Rectification

The second part of this lab utilized an uncorrected image of Sierra Leone Africa (Figure 5), and the reference Sierra Leone image (Figure 6) of this area, to correct the incorrect image. As oppose to the first part, this process involved 2 satellite images, rather than a map. A Third Order Polynomial Transformation was used with a minimum of 10 GCP points, but 12 GCP points in order to more accurately correct the image. As previously done, both images were brought into the Control Points tool (Figure 7), however in this part, a bilinear method was used rather than the nearest neighbor method. Bilinear method is more accurate than nearest neighbor because it uses the color value of the 4 surrounding pixels and averages it, where the nearest neighbor method just repositions the single pixel and uses the value from the pixel that is predominantly in the resampled image. Once both images were brought in, 12 GCP points were placed (Figure 8), and the target error was to be below 1, and ideally below 0.5 making the image more accurate.
 Figure 5: This is a distorted Satellite image fro Sierra Leone Africa that will be corrected.
Figure 6: This is the reference satellite image that will be used to correct the Sierra Leone Africa distorted image.
Figure 7: This image shows the 12 control points that were used to correct the Sierra Leone Africa Image.

Results

Part 1: Image-to-Map Rectification

The result of using a first order polynomial transformation with a nearest neighbor method created a corrected satellite image of the Chicago Illinois area (Figure 8). This image is more accurate than the original image in Figure 1 because the distortion is removed, and the image is more closely related to the topographic map of the area in figure 2. The Spatial error is also below 2 which is shown in figure 4 which means the GCP points are more accurate and closely related, which ensures less distortion in the corrected image.
Figure 8: This is the corrected Chicago area image.

Part 2: Image-to-Image Rectification

Through the use of the third order polynomial transformation with the bilineal method, the image of Sierra Leon was corrected using 12 GCP points, and a error goal of below 1. this image was a little more difficult to correct, because of the lack of distinct points in each image that are easily identifiable. In my first attempt at correcting the image, the spatial error was below 1 but the GCP points were clearly not done right because the image ended up horribly distorted (Figure 9). When GCP points are correctly aligned, and the Spatial error is below 1, a more accurate and correct image is produced (Figure 10).

Figure 9: This is what happens to an image with GCP points that are not correctly done.
Figure 10: This Image is the correct version of a properly geometrical corrected image of Sierra Leone Africa.

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.   

Comments

Post a Comment

Popular posts from this blog

Lab 8 Spectral Signature analysis and Resource Monitoring

Image
Lab 8 Spectral Signature analysis and Resource Monitoring Goal The main goal of this lab is to interpret spectral reflectance of various Earth surfaces and near the earths surface from satellite images, and to perform monitoring of surface features. In this lab we will collect spectral signatures from remotely sensed images, and create maps showing Different spectral signatures. Methods Part 1: Spectral Signature Analysis First spectral signatures were gathered by finding areas in the image that contained a surface feature we were looking for, and an area of interest was put in the area, and the spectral signature was recorded. This was done for 12 features some of them being Crops, Dry and Wet soil, standing water, moving water, and some more. Part 2: Resource Monitoring Section 1: Vegetation and Health Monitoring IN this section of part 2 vegetation was the focus, and a NDVI or normalized difference vegetation index was produced for Eau Claire and Chippewa Counties. This in...
Read more

Lab 7: Photogrammetry

Image
Lab 7: Photogrammetry Goal and Background The main goal of this lab is to develop skills in photogrammetric tasks on aerial photos and satellite images. Specifically the goal is to understand the mathematics behind calculating photographic scales, measuring areas and perimeters of features, and calculating relief displacement. In addition the lab is intended to introduce stereoscopy and performing orthorectification on satellite images. After the completion of this lab we will have the ability to perform diverse photogrammetric tasks. There are three parts to this lab: part 1-scales, measurements and relief displacement, part 2-stereoscopy, and part 3- orthorectification. Part 1, Section 1: Calculating scale of nearly vertical aerial photographs. Part 1, section 2: Measuring areas of features on a photograph. Part 1, section 3: Calculating relief displacement from object height. Part 2, Section 1: Created an anaglyph image with the use of a Digital Elevation Model (DEM). ...
Read more

Lab-5 LiDAR Remote Sensing

Image
Goal and Background The goal of this lab is to gain basic LiDAR skills through the use of ArcGIS. Some Specific goals were processing and retrieving surface and terrain models. The creation of an intensity image and other point cloud images. The LiDAR data used is in LAS file format which is easily manipulated through the use of ArcGIS. Methods Part 1: Point cloud visualization in Erdas Imagine The first part of this lab Erdas Imagine was used to input point clouds and visualize the LiDAR imagery. This was done by opening each individual LiDAR point cloud file in .las format and this gives the ability to view the LiDar image as a whole (Figure 1).  Figure 1: This image shows the importing process of LAS files into Erdas Imagine.  Next we brought these same files into ArcMap where it is easier to manipulate Lidar imagery. This was done by creating an LAS data file where LAS point cloud imagery was added of Eau Claire county. This data appears in a bunch of tile...
Read more