Melissa Hackenmueller
Geog 338
Lab
4: Miscellaneous Image Functions
Lab
Goal
Lab 4
had seven distinct learning goals that stem from various image functions that
are important in the analysis of images. The first objective was to use image
subset techniques to delineate a region of interest from a larger image. The
second was using pen sharpen to adjust the spatial resolution of an image. The
next objective was to learn some basic radiometric enhancement techniques, such
as haze reduction, to enhance image analysis skills. The fourth goal was to
enable the Google Earth feature of ERDAS Imagine to use as an aid in analysis.
The fifth objective was to understand the basic tools of resampling images. The
sixth goal of lab 4 was to use two image mosaicking procedures. The final goal
was to introduce binary change detection and its uses in image analysis.
Methodology
Part
1:
The goal of part one was to learn how to use image subset. An image subset is
creating a delineation of a region of interest from a larger satellite image
scene. The first method used was an inquire box. Once the original image
(eau_claire_2011.img) was imported I selected the inquire box from raster tools
and then I adjusted the box to the desired Eau Claire area. Lastly, I used the
tool subset and chip to create a permanent subset image. The major disadvantage
of using the inquire box method, is that only rectangles can be made. The next
method, area of interest, allows you to create a subset of any shape. The first
step of the area of interest method was to add the desired image
(eau_claire_2011.img) and then add the county shapefile (ec_cpw_cts.shp) on top
of the image. Next, I selected both counties from the shapefile and saved them
as an area of interest. Then I used the subset and chip tool again and clicked
on AOI to tell the tool to use the area of interest file that I just created.
This then created a new image, but now the image was the shape of the desired
counties.
Part
2:
The objective of part two was to use the pen sharpen tool to improve the
spatial resolution of an image. The first step I did was open the desired image
(ec_cpw_2000.img) and then the panchromatic image of the same area
(ec_cpw_2000pan.img) into a second viewer. Then I used the resolution merge
tool from the pen sharpen section of the raster tools. The resolution merge
window then popped up and I entered the correct inputs and output, then
selected multiplicative for the method, and then selected nearest neighbor from
the resampling techniques. I then ran the model to create a new image with a
higher spatial resolution.
Part
3:
The objective of part three was to learn simple radiometric enhancement
techniques. Step one was to open the ec_claire_2007.img and then I used the
haze reduction tool from the radiometric section of raster tools. I then
entered the correct inputs and outputs and ran the haze reduction model. This
created a new image that improved the scattering that was present on the
original image.
Part
4:
The goal of part four was to learn how to connect Google Earth with ERDAS
Imagine and its benefits. The first step I did for this part was importing the
eau_claire_2011.img. I then choose connect to Google Earth on the Google Earth
tab. A Google Earth Interface opened and I choose match GE to view on ERDAS
Imagine to match Google Earth with my Eau Claire image. I then sync GE to view
to link the two images. Now, I could zoom in on one image and the other would
automatically follow.
Part
5:
The objective of part five was to resample images. Resampling is the process of
changing the pixel size on an image. I started this section by importing the
eau_claire_2011.img into the viewer; then I used the resample pixel size tool
from the spatial tab of raster tools to bring up the resample window. I entered
the correct input and outputs and chose the nearest neighbor method and changed
the output size to 15x15 meters in both the x and y direction. This created a
new image with a pixel size of 15x15 meters instead of 30x30 meters. There is
multiple types of resampling methods, so I went back through the steps above
but chose the method bilinear interpolation in the resampling window and
created another image using this method.
Part
6:
The goal of part 6 was image mosaicking. This means the combining of two
satellite images. As usual, the first step was to go to the add image folder
but this time the settings need to be changed to multiple images in virtual
mosaic in the multiple tab. Then the first image eau_claire_1995p26r29.img can
be added and then the same procedure followed to input eau_claire_1995_p26r29.img.
Next, I used a simple method of image mosaicking, Mosaic Express. In the Mosaic
Express window I added both of my images that I currently have open in my
viewer, making sure to add the eau_claire_1995p26r29.img first and ran the
model. Next, I ran another version of image mosaicking called MosaicPro. First,
I added the images to the viewer as I did for the Mosaic Express method above;
then I chose MosaicPro tool from the raster tools. In the MosaicPro window I added the images,
color corrected, set my output, set my overlap function, and then ran the
model.
Part 7: The goal of this
final part of lab four was to learn about image differencing or binary change
detection. This is a feature that maps the temporal differences of an area. The
first step is to open ec_envs1991.img in one viewer and ec_envs2011 in a second
viewer. Next, I went to the two image functions tool in the functions section
of the raster tools. I then input both of the images above into the two input
operators interface. I changed the operator from addition to subtraction,
changed the layer for the inputs just to layer 4, and then I ran the model. I
then observed the metadata of this new image and found the upper and lower
change threshold using the mean and standard deviation. This gave me a rough
idea of the amount of change but I needed a more precise number. Therefore, I
started a model by going to model maker in the toolbox. I made the first model
(Figure 1) so that there would be no negative values of change as there was in
my histogram before. Then I created a second model (Figure 2) to map these
areas of change. This change map was difficult to view in ERDAS Imagine so I
used ArcMap to create a proper, readable map.
 |
| Figure 1 |
 |
| Figure 2 |
Results
Part
1:
Figure 3 is the result I got from using the inquire box tool to create a subset
image. Figure 4 is the image I got from using the area of interest method to
create a subset image.
 |
| Figure 3 |
 |
| Figure 4 |
Part
2:
Figure 5 is the comparison of the original ec_cpw_2000 image on the left and
the pen sharpened image on the right. Notice the obvious improvement in
contrast in the pen sharpened image.
 |
| Figure 5 Part 3: Figure 6 is the comparison of the original eau_claire_2007.img on the left and the improved haze reduction image on the right. |
 |
| Figure 6
Part
5:
I noticed a more apparent change in the bilinear interpolation method so Figure
7 is the original eau_claire_2011.img on the left and the resampled image on
the right. The images are zoomed in to show the difference in pixel size.
|
 |
| Figure 7
Part
6:
Figure 8 is the image mosaic I created using the Mosaic Express model. This
model doesn’t do a good job at providing a seamless image but it is simple and
quick. Figure 9 is the image I created using the MosaicPro model. This image is
much more seamless than Figure 6.
Figure 8
Figure 9
|
Part 7: Figure 10 is a histogram showing roughly how much change has occurred between 1991 and 2011 in the Eau Claire region. Figure 11 is the change mapped in ArcMap.
 
Figure 10 Figure 11
Sources
Eau Claire Remote Sensing Imagery, Earth Resources Observation and Science Center, United States Geological Survey.
Mastering ArcGIS 6th edition Dataset by Maribeth Price, McGraw Hill. 2014.
|
