Lab 1: Miscellaneous Image Functions

Goal and Background:

This lab was a laboratory exercise that’s goal was to introduce new remote sensing techniques.  There were seven parts to this lab and in each part a new remote sensing technique was learned and practiced.

• Part 1 focused on delineating study areas from a larger satellite images. 
• Part 2 demonstrated how spatial resolution could be enhanced for image interpretation purposes. 
• Part 3 gave training in radiometric enhancement techniques in visual images.   
• Part 4 worked with linking Google Earth to satellite images to obtain additional information useful in image interpretation.
• Part 5 introduced various tools of resampling satellite images.
• Part 6 practiced image mosaicking techniques. 
• Part 7 practiced binary change detection by using graphic modeling.   

Methods:

In all sections of this lab Erdas Imagine was the image processing software used in all of the following parts of the lab.  I used the geography program ArcMap to create the map in Part 7: Section 2.  All images came from the Earth Resources Observation and Science Center, United States Geological Survey. 

• Part 1: Part one was divided into two subsections.
• Section 1: Section one worked with subsetting an image by using an Inquire box.  To execute this I started by opening the image eau_claire_2011.img image in Erdas Imagine and activated the raster tools.  A right click on the image allows for the Inquire Box option to appear.  A white box appears on the image and then by holding down the left mouse cursor I moved the box to the proper area over the Eau Claire and Chippewa area.   I also made the box larger by dragging the corner of the box to create a larger square of the size I desired, able to encompass all of Eau Claire and Chippewa.  Once the box was set in the desired location and size I clicked apply on the Inquire box.  I clicked on raster tools and choose Subset and Chip and Create Subset Image tools from the drop down menu.  Once in the window I properly set location for proper saving and naming of the image and chose the From Inquire Box option in this window allowing for the coordinates of the image to also occupy the space in the subsetted area.  I left all of the other parameters and clicked ok.  After the completion of the model run I dismissed and closed out the processing list window.  The new subsetted image was available in the location I set for it to save to.       
• Section 2: Section two taught how to subset a given area by using an interest shapefile.  I began by opening the eau_claire_2011.img in Erdas Imagine.  I also added the ec_cpw_cts.shp shapefile to the viewer.  In order to this I had to change the format of the file from .img to .shp in the File of type drop down menu in the Select Layer to Add window that is used when adding images to a viewer.  The shape file was overlaid on the Eau Claire image in the viewer in a blue color.  I selected the Eau Claire and Chippewa counties contained in the shapefile one at a time by holding down the shift key.  The counties became selected when they became yellow in color.  I then clicked on Home in the toolbar and then paste from selected object on the Home toolbar.  Next, I clicked on File, Save As-AOI Layer As and saved in a designated fold location.  I then clicked on Raster and then Subset and Chip.  I set the location to where I wanted the image to be saved to and clicked on the AOI button on the bottom of the Subset window.  I clicked ok and brought the newly subsetted image into the window from the designated area it was saved to.  
• Part 2: Part two worked with the tool of image fusion.  I began by opening image ec_cpw_2000.img in a viewer in Erdas Imagine.  I then opened a second viewer where I brought in the panchromatic version of this image, ec_cwp_2000pan.img.  The 15 meter panchromatic image was used to pan sharpen the 30 meter reflective image.  I completed this by clicking on raster to activate the raster tools in the main toolbar.  I then clicked on the Pan Sharpen icon and chose Resolution merge from the dropdown menu.  In the Resolution Merge popup window I entered the panchromatic image as the High Resolution Input File and entered the reflective image as the Multispectral Input File and entered a place for the output image to go in Output File.  Under the Method heading I selected Multiplicative and under the Resampling Techniques I selected Nearest Neighbor.  I left all other default setting alone and clicked ok to run the model.  After the model finished running I clicked dismiss and clear.  I then brought the finished image into the view from the location I designated for it to save to.                
• Part 3:  Part three worked with the radiometric technique of haze reduction.  I started by opening image eau_claire_2007.img in Erdas Imagine and clicked on raster to activate the raster tools.  I clicked on Radiometric and Haze Reduction.  I inputted this selected image for the input file and created a place to save the image to.  I accepted all of the default parameters and clicked ok.  My output image was found in the location I set for it to export to.  
• Part 4: In part four I used Google Earth to help as an image identification key by linking the Erdas Imagine viewer with a Google Earth window.  I started by loading the eau_claire_2011.img into Erdas Imagine and clicked on the Google Earth icon under the Google Earth heading in the main toolbar of the Erdas Imagine interface.  I then clicked on the Connect to Google Earth icon and the Sync GE to View icon.  I then was able to have both the Google Earth window and Erdas Imagine viewer synced in view and allowed for deeper interpretation of the image.  
• Part 5: In part five I worked with the process of resampling.  To start I brought the image eau_claire_2011.img into an Erdas Imagine viewer.  I then opened the Image Metadata for the image.  To take note of the images pixel size, 30x30 meters.  I then closed that window and clicked on Raster in the main toolbar and chose Spatial and Resample Pixel Size from the dropdown menu.  This caused the Resample window to open and I entered in the image as the input image and selected a location to output my image to.  For a Resample Method I chose Nearest Neighbor from the dropdown menu and changed the Output Cell Size from 30x30 meters to 15x15 meters.  I also checked the box for Square Cells before clicking ok.  I repeated this process again but when choosing a Resample Method I chose Bilinear Interpolation.           
• Part 6: Part six was divided into two subsections.  Both of the two subsections were started by loading image eau_claire_1995p26r29 into Erdas Imagine, but before loading the image into the viewer in the Select Layers to Add window under the Multiple tab, Multiple Images in Virtual Mosaic had so be selected and in the Raster Options tab Background Transparent needed to be checked.  I then clicked ok to load the image into the viewer.  These same steps were taken for bring in the second image eau_claire_1995p25r19.img.           
• Section 1: Section one worked with image mosaicking by using Mosaic Express.  With both images loaded in the viewer I started by clicking on the Raster tab in the main toolbar and then selected the Mosaic tools and then selected Mosaic Express from the dropdown menu.  I fist loaded in the image eau_claire_1995p25r29.img before loading in the image eau_claire_1995p26r29.img into the Input Mosaic Express window.  I accepted all default parameters and in the Root Name box selected the location I desired for my image to be output to.  I clicked finish to run the model and once the run was complete I clicked dismiss and brought the finished image into the view from the location I designated for it to save to.       
• Section 2: Section two worked with image mosaicking with the use of MosaicPro.   With both images loaded in the viewer I started by clicking on the Raster tab in the main toolbar and then selected the Mosaic tools and then selected Mosaic Pro.  I then clicked the tool in the toolbar of the newly opened window that added images.  I clicked to highlight the image eau_claire1995p25r29.img and then clicked the Image Area Options and clicked Compute Active Area and clicked the set button.  I left all of the default parameters and clicked ok to add the image.  I repeated the above actions to also add the image eau_claire_1995p26r29.img. To synchronize the radiometric properties of the two images I clicked the Color Corrections Icon and checked the box for Use Histogram Matching.  I clicked the set button and selected Overlap Areas in the top of the toolbar on the right.  I left the default settings for Histogram Matching and clicked ok in the window.  I clicked ok in the Color Corrections window to make the updates.  In the MosaicPro toolbar I clicked the Set Output Options Dialog icon to open the Output Image Options window.  I accepted all default parameter and closed the window.  I clicked on the Set Overlap Function icon to open the Set Overlap Function window and accepted the default method of Overlay.  I clicked the Process button in the MosaicPro window and Run Mosaic.  I then set the location of output and named the output image.  At the end of the model run I clicked dismiss.  I then brought the newly created image into the Erdas Imagine Viewer.                  
• Part 7: Part seven was divided into two subsections.
• Section 1: Section one worked with analyzing the pixels that changed between two images.  To start in Erdas Imagine I opened two viewers.  In the first viewer I added the image ec_envs1991.img and in viewer two I added the image ec_envs2011.img.  I fit both imaged to view and synched the viewers. I clicked on Raster on the main toolbar to activate the raster processing tools.  I then selected Functions and then chose Two Image Functions option to access the Two Input Operators selection.  For Input File #1 I input the image ec_envs2011.img image and input image ec_envs1991.img image into Input File #2.  I selected a location for my output image file.  Under the Operator dropdown menu I changed it from + to -.  In the drop down menus under Input File #1 and Input File #2 in the two input Operators window in the layer dropdown menu I changed it from all to 4.  I clicked ok to run the image differencing tool.  At the end of the model run I clicked dismiss and brought in the newly created image.  I then estimated the change-no change thresholds using the histograms in Image Metadata.  To do this I opened the new image's Image Metadata and clicked on the Histogram option.  In using the Image Metadata under the General tab I took the standard deviation and multiplied it by 1.5 and added it to the mean. Using these numbers I obtained the point for the upper point for the change-no change threshold, 74.26.  I repeated the same steps to obtain the lower point for the change-no change threshold, -22.06.                        
• Section 2: Section two worked with the mapping of the changing pixels from an image differencing image by using spatial modeler.  I opened Erdas Imagine with a blank viewer. I clicked the Toolbox tab on the main toolbar and selected the Model Maker tool and then Model Maker from the dropdown menu.  The new model menu opened up and I used the selection tool to add two raster objects next to each other in the window.  I then added a function object below the two raster objects in the window.  I placed a third raster object below the function object.  I connected all of the objects using connectors.  I uploaded the 2011 NIR band image ec_envs_2011_b4.img to the first raster object.  I brought the 1991 NIR band ec_envs_1991b4.img into the second raster object.  For the function object I entered an equation subtracting the 1991 image from the 2011 image and added the constant, $n1_ec_envs_2011_b4 - $n2_ec_envs_1991b4 + 127.  In the last raster object I added the output location and named this new image.  I then clicked the button in the toolbar of the window that runs the model.  I opened the new image in the viewer in Erdas Imagine and opened image's Image Metadata and selected the histogram option.  I determined the new threshold by using the mean and adding three times the standard deviation, 202.15.  I opened a new Model Maker tool and added a raster object, a function object right below it and another raster object right below the function object.  I then connected the three objects with connectors.  I loaded the image that was just created in the first raster object and in the function object I changed the function from Analysis to Conditional in the dropdown menu in the upper right and selected an Either IF OR function in the window below that.   I then wrote out the following equation, EITHER 1 IF ($n1_dif_ec_91>202.15) OR 0 OTHERWISE.  In the last raster object I specified a place and name for the new image.  I then ran the model and brought the new image into the viewer.  I then opened ArcMap and opened the image that was just created and the ec_envs_1991by.img in the viewer.  For the recently created image I set the data to no data and color to no color so just the changed portions of the map were showing.  I added a legend, scale, and north arrow to the map before exporting the image.         

Results:

• Part 1: Image subsetting (creating an area of interest AOI) of a study area 
• Section 1: Using an Inquire box to subset

Figure 1

Figure 1 is the image produced by using an inquire box to subset the Eau Claire and Chippewa area.

• Section 2: Using an area of interest shapefile to subset

Figure 2

Figure 2 is the image produced by using an area of interest shapefile to subset the Eau Claire and Chippewa area.

• Part 2: Image Fusion: Pan Sharpen

Figure 3

Figure 3 is the image produced by using the image fusion tool called Pan Sharpen.  The colors in a Pan Sharpened image are more realistic looking than the colors in a reflective image.  Pan-sharpened images are clearer and are better images to be used for image interpretation purposes.

• Part 3: Radiometric enhancement techniques: Haze Reduction

Figure 4

Figure 4 is the image produced by using the enhancement technique Haze Reduction.  This image is much clearer than the original image from before the haze reduction technique was applied.  In this image the features are sharper and the colors are brighter.  The white hue that appeared on the reflexive image, see Figure 2, is no longer present in this figure.  This image is easier to use for image interpretation than the original with the white haze.       

• Part 4: Linking a satellite image in a Erdas Imagine viewer to Google Earth

Figure 5

Figure 6

Figures 5 and 6 both show screen shots of Google Earth linked to the Erdas Imagine viewer.  Google Earth can be used as an image interpretation key, it can aid to the interpretation of an aerial photograph or satellite image.  Google Earth is a selective key because, it contains images with supporting text.  This is shown in Figure 6.  As the satellite image is loosing clarity the Google Earth image still shows a clear image and gives information about the features in that area in the band at the bottom of the viewer.  

• Part 5: Resampling

Figure 7

Figure 8

Figures 7 and 8 both show the image output I received from using the Nearest Neighbor resampling method featured in the images on the right side.  The images on the left side are the original images form before the resampling technique was used.  From Figure 7 you cannot visually detect much difference from the original image and the resampled one.  In figure 8 with the images more zoomed in you can notice that the resampled image contains larger pixels and lost its image clarity faster than the original image.   

Figure 9

Figure 10

Figures 9 and 10 both show the image output I received from using the Bilinear Interpolation resampling method featured in the images on the right side.  The images on the left side are the original images from before the resampling technique was used.  From Figure 9 you cannot visually detect much of a difference between the original image and the resampled one.  In figure 10 with the images more zoomed in you can notice that the resampled image has smaller pixels and can keep its clarity longer than the original image. 

• Part 6: Image Mosaicking
• Section 1: Image Mosaicking with the use of Mosaic Express

Figure 11

Figure 11 is the image produced by using the image mosaicking tool Mosaic Express used to combine two images that are overlapping.  

• Section 2: Image mosaicking with the use of MosaicPro 

Figure 12

Figure 12 is the image produced by using the image mosaicking tool MosaicPro used to combine two images that are overlapping.  

• Part 7: Binary change detection (image differencing)
• Section 1: Creating a difference image

Figure 13

Figure 13 is an image of the histogram produced by the differencing of two images.  The lower point and the upper point of the change-no change threshold are displayed on the histogram.  

• Section 2: Mapping change pixels in difference image using spatial modeler

Figure 14

Figure 14 is the map that was created to display the areas of the Eau Claire that have changed in spatial distribution over the last 20 years.  These areas of change have been identified by using a spatial modeler.  The areas with the greatest amounts of change are located in the areas of land along rivers where there are more urban and populated land areas.  

Sources:

"Home page- Earth Resources Observation and Science (EROS) Center." Home Page | Earth

Resources Observation and Science (EROS) Center. N.p., n.d. Web. 29 Mar. 2016. <http://eros.usgs.gov/>.

" Earth." Google Earth. N.p., n.d. Web. 30 Mar. 2016. <http://earth.google.com/>.